TWI229825B - Direct mode motion vector calculation method for B picture - Google Patents

Abstract

A method for calculating direct mode motion vectors of a B (Bi-predictive) picture in a moving picture coding system to extract the direct mode motion vectors of the B picture. When a list 1 reference picture for direct mode temporally precedes or follows the B picture, any one of motion vectors of a co-located block in the list 1 reference picture is determined as a motion vector for derivation of the direct mode motion vectors regardless of modes (a list 0 mode and/or a list 1 mode) of the motion vectors of the co-located block. The determined motion vector is scaled to derive a list 0 motion vector MVF and a list 1 motion vector MVB. Further, a reference picture pointed by the determined motion vector is determined as a list 0 reference picture for direct mode.

Description

1229825 五、發明說明（1) [發明所屬之技術領域] 本發明有關一種動晝編碼方法，尤其有關次代 ::定義的B圖像(雙預測圖像)的直接模式運動向旦量導出1229825 V. Description of the invention (1) [Technical field to which the invention belongs] The present invention relates to a moving day coding method, and in particular, it relates to the next generation :: defined B-picture (bi-predictive picture) direct mode motion derived to the denier

[先前技術J 象Λ五種預測模式類型，諸如正向模式、反向模 式、直接模式、’在内’模式。在正向模式、反 識二為莫ί =量的方向可從模式名稱辨 以時間7Γ私电％中已括了方向貧訊。在直接模式中，是 持運‘連：：ί f為基礎，'亦即’以兩相鄰圖像間經常維 ί = ;::基：運LI圖像内的共位置區塊的運動 效率，因為並不；送J二1。直接模式的優點在於編碼 +得迗運動貧訊給解碼器。 part 10，盆在戈動晝壓縮技術中，例如H· 264或MPEG-4 圖像使用，因為它可儲，ί特破在於，6圖像可當作基準 圖像之特徵進而包括.二 基準圖像緩衝區内。此種β ，list 0,模式、，u •匕=五種預測模式類型，諸如 ，在内，模式。 杈式、雙預測模式、直接模式' ’ 1 i s t 0 ’模式類似習用 如基準圖像索引與運動向旦：模式，#中之運動資訊，諸 mvd_10指示之。，u 二差，係分別用ref_idx_l〇及 中之運動資訊，諸如類似習用的反向模式’其 用ref」dx_U及mvd广二圖像索引與運動向量差’係分別 - 1 ί日不之。雙預測模式有二個基準圖 第5頁 1229825 五、發明說明（2) Ϊ:: ί ΐ0!間上而言都可位於b圖像之前或之後；或 此情；時:位於㈣像之前或之後。在 、用 ref — ldx」〇、ref 一 idx 一 11 與 mvd—1〇、 每：個基準圖像索引與二個運動向量差，且 數（pot) Λ / 像料計數（pgg) f料。®像順序計 要U尸ϋΐ)/枓是一種時間位置資訊。 、軍ί 3 ?式中’藉由選擇空間技術與時間技術之-來獲得 運動向1。空間直接模式技術是從一欲 = ㈣塊導出llst Gmlst i基準圖像索引與運動^各^ 日守間直接模式技術中，藉由測量一直接模式用1 ist J基準 圖，内共位置區塊的list 〇運動向量，導出__list 〇 ^動 向量MVF與一iist 1運動向量MVB。直接模式用!芙 圖像類似習式的B圖像。其中，直接模式用丨ist j基準&圖 像的list 1預測用索引為〇，而直接模式用Ust 〇基準圖 像是直接模式用list 1基準圖像内共位置區塊運動向量 指向的1 i s t 〇基準圖像。 里 圖1 a至1 c分別顯示呈IBBBP形式B圖像的丨i st 〇預測用預气 索引、list 1預測用預設索引、及直接模式用list 1基2 圖像’其中可用之1 i st 0與1 i st 1基準圖像數（或一短^ 緩衝區之大小）為6。其中，1 i s t 0預測用預設索弓丨與1 ^ 1預測用預設索引都視一先前解碼之基準圖像的輪1 S t (或POC值）而定，而與解碼順序無關。圖1中，所有B β 都使用一時間上在後的Ρ圖像作為直接模式用丨i st〗f像 圖像。 '"準[Prior art J looks like five types of prediction modes, such as forward mode, reverse mode, direct mode, and 'in' mode. In the forward mode and the reverse direction, the direction of the quantity can be discerned from the mode name. The time 7Γ private electricity% has included the lean direction. In the direct mode, it is based on the operation of ':: ί f', that is, 'that' is based on the constant maintenance between two adjacent images. , Because not; send J 2: 1. The advantage of the direct mode is that it encodes + succeeds the motion to the decoder. Part 10 is used in the GD compression technology, such as H.264 or MPEG-4 image, because it can be stored. It is particularly broken that the 6 image can be used as the feature of the reference image and then includes. Image buffer. This β, list 0, mode, u u = five types of prediction modes, such as, mode. Bifurcation, bi-prediction mode, direct mode '′ 1 i s t 0 ’mode is similar to conventional applications, such as the reference image index and motion direction: mode, motion information in #, which are indicated by mvd_10. The difference between u and u is based on the motion information of ref_idx_l0 and, respectively, such as the reverse pattern similar to the conventional one, which uses ref "dx_U and mvd. The bi-prediction model has two reference maps. Page 5 1229825 V. Description of the invention (2) Ϊ :: ί ΐ0! Can be located before or after the b image; or in this case; hour: before the image or after that. At, use ref — ldx ”0, ref — idx — 11 and mvd — 10, each: the difference between the reference image index and the two motion vectors, and the number (pot) Λ / image count (pgg) f material. ®Sequence counts like U ϋΐ) / ϋΐ is a kind of time position information. ， 军 ί 3? In the style, ‘by choosing one of the space technology and the time technology, the movement direction 1 is obtained. The spatial direct mode technique is to derive the llst Gmlst i reference image index and motion from a desire = ㈣ block. In the day-to-day direct mode technology, a 1 ist J reference map is used to measure a direct mode, with a total of internal block locations. The list 〇 motion vector is derived from the __list 〇 ^ motion vector MVF and an iist 1 motion vector MVB. The direct mode uses the! F image similar to the conventional B image. Among them, the index for prediction of the list 1 of the direct mode ist j reference & image is 0, and the reference of the direct mode Ust 0 is the 1 pointed to by the motion vector of the co-location block in the list 1 reference image of the direct mode. ist 〇 benchmark image. Figures 1a to 1c respectively show the i-st st of the B-picture in the form of IBBBP. Pre-expected index for prediction, preset index for list 1 prediction, and list 1-based 2 images for direct mode. The number of st 0 and 1 i st 1 reference images (or the size of a short ^ buffer) is 6. Among them, the preset index for 1 s t 0 prediction and the preset index for 1 ^ 1 prediction both depend on the round 1 S t (or POC value) of a previously decoded reference image, regardless of the decoding order. In FIG. 1, all B β use a time-lagging P image as a direct mode image. '" quasi

圖4 a至4 h顯示， 一 B圖像之前時 有的幾種模式。 1229825 五、發明說明（3) 圖2a至2c分別顯示呈ΙβΒΒ形式B圖像的1 i st 〇預測用預設 索引、1 i s t 1預測用預設索引、及直接模式用1丨s t 1基準 圖像’其中只使用B圖像。在圖2a中，當欲編碼之B圖像為 B8時’時間上在前且丨ist }索引為〇之B5，是直接模式用 1 i s t 1基準圖像。如圖2 b所示，後續欲解碼之β 7，其直接 模式用list 1基準圖像是時間上在後之β8。最後，如圖& 所示，後續欲解碼之β 9，其直接模式用1丨s t }基準圖像是 時間上在前之β 7。 圖13至2。所示，直接模式用ust 1基準圖像可為 寻間上在欲編碼B圖像之後的P圖像或B圖像，或是時間上 在欲編碼B圖像之前的b圖像。 當直接模式用Ust 1基準圖像為時間上在 圖像之後日守，在1 i st i基準圖像内的共位置 有的幾種模式。在此情況時，由於1 i s t丨其.隹‘ 、月匕 P圖像或-B圖像’所以其内的位 ；準：像可為- 向量或為，在内，桓4。a处^ 置£塊有—或二個運動 MPEG-4 Part 10 二許= 技f ’ 諸如H.264 或 =产Hs"預測用的索 基準圖像可以恰在—_像之後 ：=l1St i 向量可朝前或朝後\ ，、位置區塊之運動 产〗· ，，# 吞竿圖像為時間 ’在list 1基準圖像内的 户仏味、口士 口1豕π的共位置區塊^ 在此情况日守，共办罢反％ aFigures 4a to 4h show several modes before a B image. 1229825 V. Description of the invention (3) Figures 2a to 2c show the 1 i st 〇 prediction index, 1 ist 1 prediction index, and the direct mode 1 st 1 reference picture in the form of I β Β B image. Like 'where only B images are used. In FIG. 2a, when the B picture to be encoded is B8, B5 whose time is ahead and whose ist} index is 0 is a 1 s t 1 reference picture for direct mode. As shown in Fig. 2b, for the subsequent β7 to be decoded, the direct mode uses the list 1 reference image to be the next β8 in time. Finally, as shown in Figure &, for the subsequent β 9 to be decoded, its direct mode uses 1 丨 s t} The reference image is β 7 which is temporally preceding. Figures 13 to 2. As shown, the ust 1 reference picture for the direct mode can be a P picture or a B picture after the B picture to be encoded in time, or a b picture before the B picture to be encoded in time. When the direct mode uses the Ust 1 reference image as the timekeeping behind the image in time, there are several modes of co-location within the 1 i st i reference image. In this case, since 1 i s t 丨 its. 隹 ′, moon d P image or -B image ′, the bit within it; quasi: the image can be-vector or be, including, 桓 4. A place ^ block has-or two motion MPEG-4 Part 10 two licenses = technology f 'such as H.264 or = production Hs " prediction reference image can be just after the _ image: = l1St i The vector can be forward or backward \ ,, and the motion product of the location block..., # # The rod image is the co-location area of the household taste and the mouth mouth 1 口 π in the reference image of time 1 in the list 1 reference image. Block ^ In this situation, day guard, strike together% a

1229825 五 '發明說明（4) ^量，為’在内’模式，如上所述。list !基準圖像血請 像之間，可能有其他基準圖像出現，所以共位置圖 運動向量可能指向時間上為前或後之方向。、位置£塊的一 1圖3a至仏可以看出，直接模式用list 1基準圖像可 種不同模式’ ϋ此，鐘於此等不 =各 法’用以計算直接模式的運動向量。…-索-種方 [發明内容] 3 ’鑑於上述問題’特此提出本發明。本發明之 =提供-種方法’用以計算次 丄在1229825 Five 'Explanation of the invention (4) The amount is in the' inside 'mode, as described above. list! Between reference images, other reference images may appear, so the co-location map motion vector may point forward or backward in time. It can be seen from Figures 3a to 仏 that the direct mode can use the reference image of list 1 to have different modes. Hereby, the clocks are not equal to each method to calculate the motion vector of the direct mode. …-索-种 方 [Summary of the Invention] 3'In view of the above problems', the present invention is hereby proposed. The present invention provides = a method to calculate the time

圖像（雙預測圖像，Bi_nrpH. +. . 1孜何'^疋義的B 運動向量。”建策P= m P1CtUre)的直接模式 技術，α祖I I 種導出B圖像直接模式運動向量之 H 升k取直接模式作為巨區塊預測模式的可 性，以便改進B圖像編碼效率。 、飞的了月匕 根據本發明一層面，豆 碼系統中B(雙預洌）圖像之吉拔“計算動晝碥 之直接模式運；Γ量二 包括以下步驟：若』 、〔及其他目❸。此方法 置區塊有-個運^ 1基準®像内的共位 ⑴ 運一:向？二 量導，像之直；模式』=幻’並從選取的運動向 步ΠίΠ:步驟：選取lls"與llst I 量，所直接模式運動向量用的運動向Image (bi-prediction image, Bi_nrpH. +.. 1 meaning B motion vector. ”Jiance P = m P1CtUre) of the direct mode technology, α ancestor II derives the direct mode motion vector of the B image In order to improve the efficiency of B-picture coding, the direct mode can be used as a huge block prediction mode in order to improve the efficiency of B-picture encoding. According to the first aspect of the present invention, the Jibba "calculates the direct mode operation of the dynamic daytime; Γ amount two includes the following steps: if", [and other objectives. In this method, the block has a common position in the reference image. ⑴ One: direction? Two vector guides, the image is straight; the mode "= magic" and from the selected motion direction Step ΠίΠ: Step: Select the amount of lls " and llst I, the motion direction of the direct mode motion vector

Ust 1 A準円f動向量應指向時間上較接近直接模式用 準圖像的圖像；若二個運動向量係指向同一基準 第s頁 1229825The Ust 1 A quasi- 円 f motion vector should point to an image that is closer to the quasi-image in direct mode in time; if the two motion vectors point to the same datum p. 1229825

圖像，則選取list 0運動向量作為、 用的運動向量；將所選取 旦 接板式運動向量 直接模式用1 i s t 0基準圖俊· 土準圖像決定為 泎、富么a曰 > 〜 冢’及’導出B圖傻夕古技上Image, the list 0 motion vector is selected as the used motion vector; the selected plate-type motion vector direct mode is used as the 1 ist 0 reference map. 'And' export B picture silly evening on ancient techniques

式運動向1。或者’以上步驟可包括以下^像之直接模 取list 0運動向量作為導出直接模式運無條件選 量，不論其時間距離如何；將list 〇運運動向 準圖像決定為直接模式用Hst〇基準圖:^旨向的基 圖像之直接模式運動向量。 ’及’導出B 根據本發明另一層面，其中提供一種 『”中B (雙預測）圖像之直接模式運動向^^出= 像之直接杈式運動向量。此方法包括以下步… ® 式用11 s t 1基準圖像内共位置區塊之運動向量中攸登f杈 何運動向量-，作為導出直接模式運動向量用的運二任 量，不論共位置區塊之運動向量其模式如何（list 〇 ° 及/或list 1模式）；將選取之運動向量所指向的基準 決定為直接模式用list 〇基準圖像；以及，計算B'"圖像之 直接模式運動向量。以往曾有人建議從一共位置區塊的 〇運動向量導出直接模式運動向量。應用此一習用方 法日守’若1 i s t 1基準圖像内的的共位置區塊只有_個1丨s t 1運動向量，所有直接模式運動向量都變成〇，因為1 ist 〇 運動向里為〇。然而，本方法可克服此一問題。” 根據本發明再一層面，其中提供一種方法，用以計算動書 編碼系統中B (雙預測）圖像之直接模式運動向量以取”出B圖 像之直接模式運動向量。此方法包括以下步驟：若直接模式 运动 向 1。 Toward 1. Alternatively, the above steps may include the following steps: directly take the list 0 motion vector as the derived direct mode unconditional selection, regardless of its time distance; determine the list movement direction quasi image as the Hst〇 reference map for the direct mode : ^ Direct mode motion vector of the base image. 'And' Derive B According to another aspect of the present invention, a direct mode motion direction of a B (bi-prediction) image in "" is provided ^^ out = a direct branch motion vector of the image. This method includes the following steps ... Use the motion vector-in the motion vector of the co-location block in the 11 st 1 reference image as the task for deriving the direct mode motion vector, regardless of the mode of the motion vector of the co-location block ( list 〇 ° and / or list 1 mode); determine the reference pointed by the selected motion vector as the direct mode list 〇 reference image; and calculate the direct mode motion vector of the B '" image. Some people have suggested in the past The direct mode motion vector is derived from the 0 motion vector of a total position block. Applying this customary method Rishou 'if the co-position block in the 1 ist 1 reference image has only _ 1 1 st 1 motion vectors, all direct modes The motion vectors all become 0, because 1 ist 〇Motion inward is 0. However, this method can overcome this problem. "According to yet another aspect of the present invention, a method is provided for calculating moving books. System B (bi-predictive) picture of the direct mode motion vectors in direct mode motion vectors to take "a B image of FIG method comprising the steps of: if the direct mode

第9頁 1229825 五、發明說明（6) 式用list 1基準圖像内的的共位置區塊只有一個list i運 動向量時，假設該共位置區塊有一零運動；將時間上恰位 於B圖像前的解碼圖像，決定為直接模式用1 i s t 〇基準圖 像；並導出B圖像之直接模式運動向量。 根據本發明再一層面，其中提供一種方法，用以計算動書 編碼系統中B (雙預測）圖像之直接模式運動向量以取出b圖 像之直接模式運動向量。此方法包括以下步驟：若直接模 式用1 i s t 1基準圖像内的的共位置區塊只有一個丨丨$七1運 動向量時，使用該共位置區塊的1 i st 1運動向量作為導出 ^接模式運動向量用的運動向量；將時間上恰位圖 前的解碼圖像，決定為直接模式用丨i st 〇基 出β圖像之直接模式運動向量。 像，並¥ 根據本發明再一層面，其中提供一種方法，用以 編碼糸統中Β(雙預測）圖像之直接模式 f = 像之直接模式運動向量。此方法包括以下動步6里以二， 式用list 1基準圖像内的的共位置區塊只個右直接模 動向量時，使用該共位置區塊的丨丨以 1運 直接模式運動向量用的運動向量；作為導出 運動向量所指向的基準圖像，決定為直2塊的list i 根據本發明再一層φ，其象中之提直運動向量。 編碼系、统中Β(雙預測）圖像之直接 法用U計算動畫 像之直接模式運動向量。此方法包括以下牛向量以取出U 碼之圖像設定為直接模式用list 1基準心驟.將最新解 測量直接模Page 9 1229825 V. Explanation of the invention (6) When the co-location block in the reference image of list 1 has only one list i motion vector, it is assumed that the co-location block has zero motion; The decoded image before the image is determined to be a 1 ist 〇 reference image for the direct mode; and the direct mode motion vector of the B image is derived. According to another aspect of the present invention, a method is provided for calculating a direct mode motion vector of a B (bi-prediction) image in a moving book encoding system to extract a direct mode motion vector of a b image. This method includes the following steps: if the direct mode uses only one co-location block in the 1 ist 1 reference image and only one $ 7 1 motion vector, use the 1 i st 1 motion vector of the co-location block as the derivation ^ The motion vector used for the motion vector of the connection pattern; the decoded image immediately before the bitmap in time is determined to be the direct pattern motion vector of the β image based on the direct mode. According to yet another aspect of the present invention, a method is provided for encoding the direct mode of a B (bi-prediction) image in the system f = direct mode motion vector of the image. This method includes the following steps 6 and 2. When the co-location block in the list 1 reference image uses only the right direct mode motion vector, the co-location block uses the direct mode motion vector in 1 The motion vector used; as the reference image pointed to by the derived motion vector, the list i is determined to be a straight 2 block. According to the present invention, another layer φ is used to straighten the motion vector. The direct method of the B (bi-prediction) image in the coding system and the system uses U to calculate the direct mode motion vector of the animation image. This method includes the following bull vector to extract the U code image and set it to the direct mode using list 1 benchmarks. The latest solution measures the direct mode

第10頁 1229825 五、發明說明（7) — 式用list 1基f圖像内共位置區塊之運動向量 ·Page 10 1229825 V. Description of the invention (7) — The motion vectors of co-located blocks in the list 1 basis f image

Hst〇運動向量MVF與Θ 出 ^ lst 1運動向量MVB ;並計曾β θ 之直接模式運動向量。在—習用方法中 1二圖像 =引為0之圖像’定義為直接模式用Ust ! t 1預測 =圖像與索引為0之圖像間有一不同圖圖像。 持索引為0之圖像的運動資訊與基準圖像資訊/ :頁維 使用記憶體。’然而，本方法可節省額外使用記憶體成頟外 根據本發明再一層面，1巾 的… 丄η, 八甲徒供種方法，用以計曾叙全 編碼糸統中Β(雙預測）圖像之直接模式運動H ^ J之直接模式運動向量。此方法包括以下步驟：若取直出接'圖 f用11 St 1基準圖像在時間上為Β圖像之前時，測量果 模式用list 1基準圖像内共位置區塊之運動向量，以 】運^量，與list lif動向量_ ;並計算b圖像 之直接杈式運動向量。 較4疋 '上步驟可包括以下步驟：若B圖像之巨區塊與 list 1 土準圖像之共位置巨區塊都為一巾貞模式，且一直接 ，式用0基準圖像在時間上為丨ist 1基準圖像之前 時’則計鼻β圖像之直接模式運動向量MVF與MVB如下： _The Hst〇 motion vector MVF and Θ give ^ lst 1 motion vector MVB; and calculate the direct mode motion vector of β θ. In the conventional method, one or two images = an image cited as 0 'is defined as a direct mode using Ust! T 1 prediction = there is a different image between the image and the image with index 0. Motion information and reference image information for images with an index of 0 /: Page dimension Use memory. 'However, this method can save additional memory usage. According to yet another aspect of the present invention, a method of ... 丄 η, Bajiatu provides a method to calculate the B (double prediction) image in the Zengxu full coding system. The direct mode motion H ^ J of the direct mode motion vector. This method includes the following steps: if the direct-connected 11 St 1 reference image is used before the B image in time, measure the motion vectors of the co-located blocks in the list 1 reference image in fruit mode, and ], And the list lif motion vector _; and calculate the direct branch motion vector of b image. Compared with the previous step, the above steps may include the following steps: If the giant block of the B image and the co-located giant block of the list 1 reference image are in a one-shot mode, and directly, use the 0 reference image in The time before the ist 1 reference image is' the direct mode motion vectors MVF and MVB of the nasal β image are as follows: _

MVF = TDB X mv / TDDMVF = TDB X mv / TDD

^ TDD) X MV / TDD 或 ζ = TDB x 256 / TDDMVF = (Z x MV + 128) >> 8 V = Z - 256 MVB = (W x MV + 128) >> g 其中’ TDB表示一現行b幀與一list 〇基準幀之間的時間距^ TDD) X MV / TDD or ζ = TDB x 256 / TDDMVF = (Z x MV + 128) > > 8 V = Z-256 MVB = (W x MV + 128) > > g where 'TDB Represents the time interval between a current b frame and a list 〇 reference frame

第11頁 1229825 五、發明說明（8) 離；TDD表不一list 1基準幀與list 〇基準幀之間的時間 距離；而MV表示直接模式用list !基準圖像内共位置區塊 之運動向量。 此外，以上步驟可包括以下步驟：圖像之巨區塊與 11 st 1基準圖像之共位置巨區塊都為幀模式，且一直接模 式用list 0基準圖像在時間上為list i基準圖像之後時， 則ό十异B圖像之直接模式運動向量μ v ρ與μ v b如下： MVF 二-TDB X MV / TDD ^ 'Page 11 1229825 V. Description of the invention (8) Leaving; TDD represents the time distance between a reference frame of list 1 and a reference frame of list 〇; and MV indicates the movement of co-location blocks in list! vector. In addition, the above steps may include the following steps: both the giant block of the image and the co-located giant block of the 11 st 1 reference image are in frame mode, and a direct mode uses the list 0 reference image in time as the list i reference After the image, the direct mode motion vectors μ v ρ and μ vb of the ten different B images are as follows: MVF II-TDB X MV / TDD ^ '

MVB = - (TDB + TDD) x MV / TDD 或 Z =- TDB X 256 / TDDMVF = (Z χ Mv + m W = Z - 256 _ = (W x MV + 128) 〉〉 8 其中，TDB表示一現行B幀與一list 離；TDD表示一list 1基準幀與丨丨“ 土其、之間的時間1 距離；而MV表示直接模式用1 i st丨其貞之間的時間 之運動向量。 1基準圖像内共位置區j 此外，以上步驟可包括以下步驟：若6圖 list 1基準圖像之共位置巨區塊都 *巨£塊| 接模式用1 i s t 0基準圖像在時間上為 #位模式’且MVB =-(TDB + TDD) x MV / TDD or Z =-TDB X 256 / TDDMVF = (Z χ Mv + m W = Z-256 _ = (W x MV + 128) 〉〉 8 where TDB means one The current B frame is separated from a list; TDD represents the time 1 distance between a reference frame of list 1 and the local time; and MV represents the motion vector of the time between 1 i st and its frame in direct mode. 1 reference In-image co-location area j In addition, the above steps may include the following steps: If the common location of the reference image of list 1 in FIG. 6 is huge, the block is huge. The 1 ist 0 reference image in connection mode is # in time. Bit pattern 'and

時，則計算一 B巾貞每一欄位i之直接模2 J圖像 MVB，i如下： 杈式運動向量MVF， MVF, i = TDB, i x MVi / TDD, i MVB, i = (TDB, i _ TDD, i) x MVi / Tdd 或Then, calculate the direct modulo 2 J image MVB of each field i of a B frame, i is as follows: branch motion vector MVF, MVF, i = TDB, ix MVi / TDD, i MVB, i = (TDB, i _ TDD, i) x MVi / Tdd or

1229825 五 8 發明說明（9) =TDB, i X 256 / TDD, IMVF, (Z x MVi + 128) >> W = Z - 256 MVB,i = (W x MVi + i28) ^ 其中，TDB，i表示一現行β欄位與__u f間距離，TDD5l表不一 llst i基準攔位 ]的 位之間的時間距離；而MVi表示一直接模t 0基準攔 攔位内共位置區塊之運動向量。 、^ 1 s t 1基準 此外’以上步驟可包括以下步驟：㈣像之巨區塊斑 11st 1基準圖像之共位置巨區塊都為一攔位 γ、 ，模式用Ust 0基準圖像在時間上為Ust 圖直 日才，則計异一 β幀每一攔位i之直 θ象之後 MVB，i如下： 之直接_式運動向量MVF，i與 MVF， i = — TDB， i X MVi / TDD， i1229825 May 8 Description of the invention (9) = TDB, i X 256 / TDD, IMVF, (Z x MVi + 128) > > W = Z-256 MVB, i = (W x MVi + i28) ^ Among them, TDB , I represents the time distance between the current β field and __u f, TDD5l represents the time distance between the bits of a llst i reference stop]; and MVi indicates a co-location block within a direct mode t 0 reference stop Motion vector. ^ 1 st 1 reference In addition, the above steps may include the following steps: The co-located giant blocks of the 11st 1 reference image of the giant block of the image are all a block γ, and the mode uses the Ust 0 reference image at time The picture above shows the direct genius of the Ust chart, and then calculates the MVB after the straight θ image of each block i of the different β frame, as follows: The direct_style motion vectors MVF, i and MVF, i = — TDB, i X MVi / TDD, i

MVB> 1 = - (TDB> 1 + TDD, i) x MVi / TDD, I 或 128) >> ;=-TDB’i x 256 / TDD，iMVF，i = (Z x MVi W = Z - 256 MVB， i = (W x MVi + 128)〉 ^中，TDB，i表示一現行“闌位與一Ust =間距離；，丄表示—】】s t】基準搁位與=閑立之間的 位之間的時間距離·，而MVi表示一直接模式用；〇基準攔 欄位内共位置區塊之運。要杈式用1 ist i基準 外，以上步驟可包括以Ϊ;驟：若B圖像… 攔位模式，而iist !基準圖像之共位巨區塊為- ^為一幀模 1229825MVB > 1 =-(TDB > 1 + TDD, i) x MVi / TDD, I or 128) > >; =-TDB'i x 256 / TDD, iMVF, i = (Z x MVi W = Z- 256 MVB, i = (W x MVi + 128)> ^, TDB, i represents a current "distance between the appendix and a Ust = ;; 丄 represents-]] st] between the reference stand and = idle The time distance between the bits, and MVi means a direct mode; 〇The reference block is used for the location of the common block. In addition to the 1 ist i benchmark, the above steps can include: 骤; if B Image ... Block mode, while the co-located giant block of the iist! Reference image is-^ is a frame mode 1229825

五、發明說明（10) 式，且一直接模式用list 0基準圖像太 >仗日等間卜炎 準圖像之前時，則計算一β幀每一攔位i之 馬llst 1基 量MVF，i與MVB，i如下： 接模式運動向V. Description of the invention (10), and when a direct mode uses list 0 as the reference image, too, it will be calculated before the image such as the day of war, etc., and then calculate the llst 1 basis for each stop i of a β frame. MVF, i and MVB, i are as follows:

MVF, i =TDB, i X MV / TDDMVF, i = TDB, i X MV / TDD

MVB， i = (TDB， i -TDD) x MV / TDD 或 128) >> 8 Z = TDB, i x 256 / TDDMVF, i = (z x Mv W = Z - 256MVB, i = (W x MV + 128) >> 其中’TDB’i表π 叫穴一11以0基準欄位之MVB, i = (TDB, i -TDD) x MV / TDD or 128) > > 8 Z = TDB, ix 256 / TDDMVF, i = (zx Mv W = Z-256MVB, i = (W x MV + 128) > > where 'TDB'i table π is called a hole 11 with 0 reference field

時間距離；TDD表示一list 1基準幀與iist 0基準巾貞之严^ 的日守間距離，而Μ V表不一直接模式用1 i s t 1基準巾貞内妓 置區塊之運動向量。 此外，以上步驟可包括以下步驟：若B圖像之巨區塊為_ 欄位模式，而1 i st 1基準圖像之共位置巨區塊為一巾貞模 式’且一直接模式用1 ist 0基準圖像在時間上為1 isj i 準圖像之後時，則計算一 B幀每一欄位i之直接模式運動： 量MVF，i與MVB，i如下： 向Temporal distance; TDD represents the day-to-day distance between a reference frame of list 1 and the strictness of the reference frame of iist 0, while MV represents the motion vector of a block in the direct frame using the 1 i s t 1 reference frame. In addition, the above steps may include the following steps: if the giant block of the B image is in the _ field mode, and the co-located giant block of the 1 i st 1 reference image is in the normal mode, and a direct mode uses 1 ist When the 0 reference image is 1 in time i isj i, after the quasi image, the direct mode motion of each field i in a B frame is calculated: the quantities MVF, i and MVB, i are as follows:

MVF, i = -TDB, i x MV / TDD MVB, i = ~ (TDB, i +TDD) x MV / TDD 或 TDB, i x 256 / TDDMVF, i = (Z x MV + 128) w - Z - 2 56MVB, i = (W x MV + 128) >> 8 8 其中’ TDB，i表示一現行B欄位與一 1 ist 0基準攔位夕„ 時間距離；TDD表示一 list 1基準幀與list 0基準巾貞之門MVF, i = -TDB, ix MV / TDD MVB, i = ~ (TDB, i + TDD) x MV / TDD or TDB, ix 256 / TDDMVF, i = (Z x MV + 128) w-Z-2 56MVB , i = (W x MV + 128) > > 8 8 where 'TDB, i represents the time distance between a current B field and a 1 ist 0 reference stop; TDD indicates a list 1 reference frame and list 0 Benchmark of Chastity

1229825 五、發明說明（11) 的時間距離；而MV表示一直接模式用list 置區塊之運動向量。 共位 此外’以上步驟可包括以下步驟：若B圖像之巨區 賴模式’而list 1基準圖像之共位置巨區塊為一攔位為 式，且一直接模式用list 〇基準圖像在時間上為丨丨“、 準圖像之前時’則從以下等式計算，之直 土 向量，娜，其中使用一list i基準鳩之攔位i = 置區塊之運動資訊，計算直接模式運動向量： ^ 1 MVF =TDB X MV1/TDD,1 MVB = (TDB -TDD, 1 ) x MV1/TDD, 1 或1229825 V. Temporal distance of invention description (11); and MV represents the motion vector of a direct mode block with list. Co-location In addition, the above steps may include the following steps: If the macro area of the B image is based on the pattern, the co-location macro block of the reference image of list 1 is a block formula, and a direct mode uses list 〇 benchmark image It is 丨 丨 in time, and the quasi-image time is calculated from the following equation, the straight soil vector, na, where a list i benchmark dove's stop i = motion information of the block is set to calculate the direct mode Motion vector: ^ 1 MVF = TDB X MV1 / TDD, 1 MVB = (TDB -TDD, 1) x MV1 / TDD, 1 or

W TDB x 256 / TDD，1MVF= (Z x MVl + 128) >〉 8 Z - 256MVB = (W x MV1 + 128) >> 8 其中’TDB表示一現行貞與一 list 〇基準傾之間的時間距 離，TDD’l表示一list 1基準幀之攔位丨與丨丨討〇基 之間的時間距離；而MV〗表示一直接模式用Hst 之欄位1内共位置區塊之運動向量。 、 此外，以上步驟可包括以下步驟：若B圖像之巨區塊 巾貞模式，而list 1基準圖像之共位置巨區塊為一°搁位模 式’且一直接模式用1 i st 0基準圖像在時間 準圖像之後時’則從以下等式計算—B賴之直接為模1 式“運1: 向量MVF與MVB，其中使用一list U準巾貞之攔位1内丘位 置區塊之運動資訊，計算直接模式運動向量： MVF = -TDB X MV1 / TDD, 1W TDB x 256 / TDD, 1MVF = (Z x MVl + 128) >> 8 Z-256MVB = (W x MV1 + 128) > > 8 where 'TDB means a current frame and a list The time distance between the two, TDD'l represents the time distance between the block 丨 and 丨 丨 base of a reference frame of list 1; and MV〗 represents the movement of the co-located blocks in column 1 of Hst in a direct mode vector. In addition, the above steps may include the following steps: if the giant block of the B image is in the normal mode, the co-located giant block of the reference image of list 1 is a 'shelving mode' and a direct mode uses 1 i st 0 The reference image after the time quasi-image is calculated from the following equation-B relies directly on the mode 1 "Run 1: Vector MVF and MVB, which uses a list U quasi-border position 1 inner mound position area Block motion information, calculate direct mode motion vector: MVF = -TDB X MV1 / TDD, 1

!229825 + TDD, 1) X MV1 / TDD, l 五、發明說明（12) mvb . 或 ;m x 256 / TDD，1MVF= (z x MV1 + 复 2 - 256MVB = (W X MVl + 128) >> 8 離=广與一llSt 0基準幢之間的時間距 之間的時門t = V ΐ準巾貞之搁位1與1 iS t 0基準棚位 丨j妁恰間距離，而MV 1表示一直接模. 干 之攔位1内共位置區塊之運動向量。、 ist 1基準幀 根據本發明另一層面，其中提供一種方 編碼系統中B(雙預測）圖像之直接模 用以叶异動真 像之直接模式運動向量。此方法向量以取出B圖 式用list 0基準圖像與list i基準 =驟·若直接模 像之後時，測量直接模式用lls"基準㊁=士都為B圖 並計算Β圖像之直接；與…^ 1心基==!:匚:::為若β:像之巨、 模式用list 0基準圖像在時間上厂幀換式，且—直接 MVF MVB 或 時，則計算β圖像之直接模式’，、、1 st 1基準圖像之後 =TDB X MV / TDD式運動向量肝卩與_如下：! 229825 + TDD, 1) X MV1 / TDD, l 5. Description of the invention (12) mvb. OR; mx 256 / TDD, 1MVF = (zx MV1 + complex 2-256MVB = (WX MVl + 128) > > 8 Time = Time gate between the time interval between Guang and a llSt 0 reference building t = V ΐ ΐ 巾 巾 贞 1 and 1 iS t 0 reference booth 丨 j 妁 distance, and MV 1 represents a Direct Modulus. The motion vector of the co-located block in Block 1. The base frame of ist 1 is according to another aspect of the present invention, in which a direct modality of a B (bi-prediction) image in a square coding system is used to discriminate the difference. The motion vector of the direct mode of the moving real image. This method vector is used to take out the B pattern and use the list 0 reference image and the list i reference = step. If after the direct image, use the lls " reference ㊁ = B for both direct mode measurements. Map and calculate the directness of the B image; and ... ^ 1 heart base == !: 匚 ::: is if β: the giant of the image, the mode uses the list 0 reference image to change the frame in time, and-direct MVF When the MVB is OR, the direct mode of the β image is calculated. After the 1st 1 reference image = TDB X MV / TDD motion vector liver 卩 and _ are as follows:

=(TDB - TDD) X MV / TDD TDB x 256 / TDDMVF - (7 v Z _ 2_ H &二;广〉> 8 1229825= (TDB-TDD) X MV / TDD TDB x 256 / TDDMVF-(7 v Z _ 2_ H &二; 广〉 > 8 1229825

其中’ TDB表不-現行B鴨與一 } i s t 〇基準賴之間的時間距 離；TDD表示一list 1基準幀與丨丨討〇基準幀之間的時間 距離；而MV表示一直接模式用list }基準圖像内共位置區 塊之運動向量。 此外’以上步驟可包括以下步驟：别圖像之巨區塊與 U st 1基準圖像之共位置巨區塊都為一幀模式，且直接模 式用Ijst 0基準圖像在時間上為Hst】基準圖像之前時， 則計异B圖像之直接模式運動向量mVF與Μνβ如 MVF = - TDB X MV / TDD 、 'Among them, 'TDB represents the time distance between the current B duck and a} ist 〇 benchmark; TDD represents the time distance between a list 1 reference frame and the reference frame; and MV represents a direct mode using list } Motion vectors of co-located blocks in the reference image. In addition, the above steps may include the following steps: Both the giant block of another image and the co-located giant block of the U st 1 reference image are in a frame mode, and the direct mode uses the Ijst 0 reference image to be Hst in time] Before the reference image, the direct mode motion vectors mVF and Μνβ of the different B images are calculated as MVF =-TDB X MV / TDD, '

(TDB + TDD) x MV / TDD W = 其中 TDB x 256 / TDDMVF = (Z χ MV + 128) » 8 - 256MVB = (W x MV + i28) 〉〉 8 離，丁DD表不一llst 1基準幀與“ J 丁门距 距離；而MV表示一直接模犬用1 . 土準幀之間的時間 塊之運動向量。 式用llst 1基準圖像内共位置區 此外，以上步驟可包括以下 # list 1基準圖像之共位置 都^圖像之巨區塊與 模式用Ust 0基準圖像在 ^為一欄位模式，且直接 時，則計算一B悄每—攔位i f 基準圖像之後 MO, i如下： 接模式運動向量MVF, i與 MVF^ = TDB}1 x MVl / 、 MVB,i = (TDB,, ^ TDD)i)；； 夕 x 評W TDD，i(TDB + TDD) x MV / TDD W = of which TDB x 256 / TDDMVF = (Z χ MV + 128) »8-256MVB = (W x MV + i28)〉 8 off, DD DD shows a different llst 1 benchmark The distance between the frame and the frame; and the MV represents the motion vector of the time block between the direct frame and the ground frame. The formula uses llst 1 to co-locate the area within the reference image. In addition, the above steps may include the following # The common position of the list 1 reference image is ^ the huge block and mode of the image. The Ust 0 reference image is in a one-column mode, and when it is direct, a B quietly is calculated after the -if reference image. MO, i is as follows: MVF, i and MVF ^ = TDB} 1 x MVl /, MVB, i = (TDB ,, ^ TDD) i); xi x W TDD, i

(Z x MVi + 128) >> 1229825 五、發明說明（14) 或 z - TDB, i X 256 / TDD, IMVF, i 8 W = Z - 256MVB, i = (W x MVi + 128) 其中，TDB，i表示一現行b攔位與一 u > 8 時間距離；TDD，i表示一list】基準 基準欄位之間的 位之間的時間距離；而MVi表示一直 st 0基準攔 攔位内共位置區塊之運動向量。 镇式用list 1基準 此外，以上步驟可包括以下步驟··若 丨基準圖像之共位置巨區塊都為ϋ之巨區塊與 杈式用list 0基準圖像在時間上為list ^ =且直接 時，則計算一β幀每一欄位i 二> Θ像之珂 MVB，i如丁·· 直接杈式運動向量MVF，i與 MVF, i = - TDB, i X MVi / TDD, i MVB,i = - (TDB, i + TDD,i) x MVi / TDD I 或 ， Z = -TDB， i X 256/TDD， iMVF， i = (Z x MVi + W = Z - 256MVB, i = (W x MVi + 128) >> 8 ) ^中，TDB，i表示一現行B欄位與一 Hst 〇基準攔位之 ^間距離；TDDj表不一list i基準攔位與丨丨以〇基準 位之間的時間距離，而MVi表示一直接模式用1丨Μ 攔位内共位置區塊之運動向量。 基準 此外，以上步驟可包括以下步驟：若B圖像之巨 攔位模式，而list 1基準圖像之共位置巨區塊為。。—鬼為“― … 巾貞模(Z x MVi + 128) > > 1229825 V. Description of the invention (14) or z-TDB, i X 256 / TDD, IMVF, i 8 W = Z-256MVB, i = (W x MVi + 128) where , TDB, i represents the time distance between a current b stop and a u > 8 time; TDD, i represents the time distance between the bits of a list] benchmark reference field; and MVi represents the st 0 benchmark stop The motion vector of the inner co-location block. In addition, the list can be used as a reference for the town type. In addition, the above steps can include the following steps: If the co-location giant blocks of the reference image are all gigantic blocks and the list 0 reference image for the branch type is in time ^ = And when it is direct, calculate each field of a β frame i 2 > Θ image ke MVB, i as Ding · · direct branch motion vector MVF, i and MVF, i =-TDB, i X MVi / TDD, i MVB, i =-(TDB, i + TDD, i) x MVi / TDD I or, Z = -TDB, i X 256 / TDD, iMVF, i = (Z x MVi + W = Z-256MVB, i = (W x MVi + 128) > > 8) ^, TDB, i represents the distance between a current B field and an Hst 〇 reference stop; TDDj indicates a list i reference stop and 丨 丨The time distance between the reference bits, and MVi represents the motion vector of a co-located block in a direct mode with 1 M block. Benchmark In addition, the above steps may include the following steps: If the B block has a huge blocking mode, the common block giant block of the list 1 reference image is. . —Ghost is "―...

12298251229825

式’且直接模式用list 0基準圖像在時間上為丨 圖像之後時，則計算一β幀每一攔位i之直接 1基準 MVF，i與MVB，i如下： 動向量 MVF, i =TDB, i X MV / TDD MVB， i = (TDB， i -TDD) x MV / TDD 或 Z = TDB, ix 25 6 / TDDMVF, i = r 7 、乙 x MV + 128、 w W = Z - 2 5 6MVB, i = (W x MV + 1 oon >> 8 1W) >> 8 其中，TDB，i表示一現行B攔位與Λ4+ /、 1 1 S t 〇基進姻从When the reference mode of the direct mode uses the list 0 reference image to be temporally after the image, the direct 1 reference MVF, i and MVB of each stop i of a β frame are calculated as follows: The motion vector MVF, i = TDB, i X MV / TDD MVB, i = (TDB, i -TDD) x MV / TDD or Z = TDB, ix 25 6 / TDDMVF, i = r 7, B x MV + 128, w W = Z-2 5 6MVB, i = (W x MV + 1 oon > > 8 1W) > > 8 where TDB, i represents a current B stop and Λ4 + /, 1 1 S t 〇

時間距離；TDD表示一list 1基準巾貞與u 千爛位之間的 的時間距離；而Μ V表示一直接模式田1 S七〇基準幀之間 置區塊之運動向量。 &丰巾貞内共位 此外，以上步驟可包括以下步驟：若3圖 棚位模式’而1 i s t 1基準圖像之共位置 、°σ塊為~ 式’且直接模式用list 0基準圖像在時為—巾貞模 圖像之前時，則計算一 B幀每一攔位i 二1 i s t 1基準 MVF，i與MVB，i如下： 接核式運動向量Temporal distance; TDD represents the time distance between a list 1 reference frame and the U bit; and MV represents a motion vector of a block placed between a direct mode field 1 S70 reference frame. In addition, the above steps may include the following steps: If the 3 image booth mode is 'and the common position of the 1 ist 1 reference image, the ° σ block is ~ type', and the direct mode uses the list 0 reference map For example, when the time is before the image of the chastity model, calculate each frame i 2 1 ist 1 reference MVF, i and MVB, i of a B frame as follows:

MVF， i = —TDB, i X MV / TDDMVF, i = —TDB, i X MV / TDD

MVB, i = - (TDB, i +TDD) x MV / TDD 或 28) >> z " -TDB, i x 256/TDDMVF, i = (Z x MV + W = Z - 256MVB, i = (W x MV + 128) >> 8 其中’TDB，i表示一現行β攔位與_list 〇基 時間距離；TDD表示一list 1基準傾與丨丨“ 〇基%情之之間MVB, i =-(TDB, i + TDD) x MV / TDD or 28) > > z " -TDB, ix 256 / TDDMVF, i = (Z x MV + W = Z-256MVB, i = ( W x MV + 128) > > 8 where 'TDB, i represents the time distance between the current beta stop and _list 〇 base; TDD represents the difference between a list 1 reference tilt and 丨 丨 〇 base% sentiment

第19頁 1229825 五、發明說明（16) 的時間距離；而Μ V表示一直接模式用1 i S t 1美進 置區塊之運動向量。 、内共位 此外，以上步驟可包括以下步驟：若B圖像之巨區土 傾模式，而list 1基準圖像之共位置巨區挣主°°塊為一 沐 尾馬一攔位揎 式，且一直接模式用1 i St 0基準圖像在時間上為丨· 模 準圖像之後時，則從以下等式計算一B幀之吉/ 1 St 1基 向iMVF與MVB，其中使用一list 1基準幀之捫%動 置區塊之運動資訊，計算直接模式運動向量： 、位 MVF =TDB X MVO / TDD,Ο / TDD,Page 19 1229825 V. The time distance of the invention description (16); and MV represents a direct mode that uses 1 i S t 1 to advance the motion vector of the block. Internal co-location In addition, the above steps may include the following steps: If the large area soil tilting mode of the B image and the co-location large area of the list 1 reference image earns a °° block is a Muwei horse and a block style And when a direct mode uses a 1 i St 0 reference image in time after the reference image, calculate the B frame of the frame / 1 St 1 basis to iMVF and MVB from the following equation, where one is used List 1 motion information of the 动% dynamic block of the reference frame, calculate the direct mode motion vector:, bit MVF = TDB X MVO / TDD, 〇 / TDD,

MVB = (tdB -TDD， 0) x MVO 或 (Z x MVO + 128)〉〉 128) >>MVB = (tdB -TDD, 0) x MVO or (Z x MVO + 128) >> 128) > >

z = TDB x 256 / TDD, 0MVF= w = Z - 256MVB = (Ψ x MVO 其中，TDB表示一現行B幀與一1 ist 〇基準鳩 離；TDD, 0表示一list 1基準幀之攔位〇盥」丨n =間距 位之間的時間距離；而MV0表示一直接模式用ust 土準攔 幀之攔位0内共位置區塊之運動向量。 卷準 此外，以上步驟可包括以下步驟：圖像之巨 =模=llst i基準圖像之共位置巨區塊為一攔鬼位為r ^ ί接模式用1iSt G基準圖像在時間上為list以進 圖像之前時，則從以下等式計算—B幀之土準 =侧，其中使用一 lls"基準巾貞之搁== 區塊之運動資訊，計算直接模式運動向量： ，、位置 MVF = -TDB X MVO / TDD,0z = TDB x 256 / TDD, 0MVF = w = Z-256MVB = (Ψ x MVO where TDB represents a current B frame and a 1 ist 〇 reference deviation; TDD, 0 represents a list 1 reference frame stop 〇 」” N = time distance between the spacing bits; and MV0 represents the motion vector of the co-location block in frame 0 of a direct pattern using ust level frame. In addition, the above steps may include the following steps: The colossal block of the common image of the reference image = modular = llst i is a block of ghosts with a position of r ^ ί. When using the 1iSt G reference image in time to list before entering the image, from the following, etc. Formula calculation—the standard of frame B = side, which uses a lls " reference frame of the frame == block motion information to calculate the direct mode motion vector: ,, position MVF = -TDB X MVO / TDD, 0

1229825 五、發明說明（17) MVB = - (TDB+TDD,0) x MVO / TDD,. 或 Z = -TDB x 256/TDD，0MVF= (Z x MVO + l28 W = Z - 256MVB = (W x MVO + 128) >> 8 >> 8 其中，TDB表示一現行B幀與一 list 0基準巾貞 離；T D D，0表示一 1 i s t 1基準幀之攔位〇與」1丨間的時間距 位之間的時間距離；而MV0表示一直接模式用1 S1 0基準攔 幀之攔位0内共位置區塊之運動向量。 1 s ΐ 1基準 根據本發明另一層面，其中提供一種方法，用、 編碼系統中Β(雙預測）圖像之直接模式運動向量^計算動畫 像之直接模式運動向量。此方法包括以下牛里以取出Β圖 間時間距離值指定一符號；測量直接模式用丨· Τ 一圖像 像内共位置區塊之運動向量，不論直接模式用/ 1基準圖 圖像與list 1基準圖像之位置如何，以導出丨丨lst 0基準 量MVF與1 i st 1運動向量MVB ;並計算B圖像夕* 0運動向 動向量。 之直接模式運1229825 V. Description of the invention (17) MVB =-(TDB + TDD, 0) x MVO / TDD ,. or Z = -TDB x 256 / TDD, 0MVF = (Z x MVO + l28 W = Z-256MVB = (W x MVO + 128) > > 8 > > 8 where TDB indicates that a current B frame is away from a list 0 reference frame; TDD, 0 indicates a 1 ist 1 reference frame block 0 and "1 丨The time distance between bits; MV0 represents the motion vector of a co-located block within frame 0 of a direct mode frame with 1 S1 0 reference frame. 1 s ΐ 1 reference according to another aspect of the invention, where Provides a method for calculating the direct mode motion vector of an animated image using the direct mode motion vector of a B (bi-prediction) image in the encoding system. This method includes the following steps to extract a time distance value between the B images and specify a symbol; Direct mode uses the motion vector of co-located blocks in an image image regardless of the position of the direct mode / 1 reference image and list 1 reference image to derive the lst 0 reference MVF and 1 i st 1 motion vector MVB; and calculate the B image * 0 motion direction motion vector. The direct mode operation

較佳是，以上步驟可包括以下步驟··若B圖像之 〇 1 i st 1基準圖像之共位置巨區塊都為一幀模式，巨區塊與 圖像之直接模式運動向量MVF與MVB如下： $ ’則計算BPreferably, the above steps may include the following steps: If the co-located giant blocks of the 0 image i st 1 reference image of the B image are all in a frame mode, the direct mode motion vectors MVF and The MVB is as follows: $ 'then calculate B

MVF = TDB x MV / TDD MVB = (TDB - TDD) x MV / TDD 或 Z = TDB x 256 / TDDMVF = (Z x MV + 128) >> W = Z - 256MVB = (W x MV + 128) >> 8MVF = TDB x MV / TDD MVB = (TDB-TDD) x MV / TDD or Z = TDB x 256 / TDDMVF = (Z x MV + 128) > > W = Z-256MVB = (W x MV + 128 ) > > 8

第21頁 1229825 五、發明說明（18) f中，TDB表示一現行β幀與一 離，若此距離係從β幀測量時，則 基準幀之間的時間距 此距離係從list 〇基準帽測量時、指定一正⑴號，若 號^表示一Us"基準續與…二其二定-負㈠ 距離，若此距離係從〗ist】基 ^丰幀之間的時間 疋負號；而MV表示一直接模式用、.里才則對其指 共位置區塊之運動向量。 、'用11^ 1基準圖像内 此外，以上步驟可包括以下步驟：Page 21 1229825 V. Description of the invention (18) f, TDB means a current β frame and a distance. If the distance is measured from the β frame, the time between the reference frames is from the list 〇 reference cap When measuring, specify a positive sign, if the number ^ means a Us " reference continued and ... two other fixed-negative distance, if this distance is from ist] the time between the base frame and the negative sign; and MV refers to a direct mode, and ri Cai refers to the motion vector of the co-location block. , 'Use 11 ^ 1 within the reference image In addition, the above steps may include the following steps:

Ust 1基準圖像之共位置巨區塊都;一圖f之巨區塊與 一B幀每一攔位i之直接模式運動向i搁位拉式，則計算 猜，i = TDB，i x MVi / TDD，i ，1 與_，1 如下： :VB’i = (TDB，i - ，〇 x MVi / 取！ Z = TDB， i x 256/tdd， iMVF i = ^Z-^BMVB,i = (WxMVl ； 128) »8 其中，TDB，i表示一現行B攔位與_ (+ )號，若此距離係從list 〇基準 私疋一正 TDD, ^ ^ ! Tst ! ^ ^ 5 "J ^ ^ 攔位之間的時間距離，若此距離係從丨i 其、 0基， 日守則對其指疋—正⑴號，若此距離係從 里 位測量日夺，則對其指定一負㈠_ 0基車欄 用list 1其i納办向认，、）虎，而MVl表示一直接模式 用list 1基準攔位内共位置區塊之運動向量。 第22頁 1229825 五、發明說明（19) 以上步驟可包括以下步驟：若B圖像之巨區 ，而list 1基準圖像之共位置巨區塊 ‘為 ^ ，„a ,、·....... & 句一幀模 此外 棚位模式’而1 i s i /土、丁叫％ <六Ί儿1 e啦塊為一 式’則計算一B幀每一欄位i之直接模式運動向量 #拉 MVB， i如下： f， 1與The Ust 1 reference image has a large number of co-located macro blocks; a block f of a picture f and a frame B each have a direct mode movement to the i-position pull-type, then the calculation is guessed, i = TDB, ix MVi / TDD, i, 1 and _, 1 are as follows:: VB'i = (TDB, i-, 〇x MVi / Take! Z = TDB, ix 256 / tdd, iMVF i = ^ Z- ^ BMVB, i = ( WxMVl; 128) »8 Among them, TDB, i represents a current B stop and _ (+), if this distance is from list 〇 benchmark private-positive TDD, ^ ^! Tst! ^ ^ 5 " J ^ ^ The time distance between the stops. If this distance is from 丨 i, 0 bases, the Japanese code refers to the 疋-positive ⑴, if this distance is measured from the inside position, a negative ㈠_ The 0-based car column uses list 1 to accept the request, and the tiger, and MV1 represents the motion vector of a common position block in the list 1 reference stop in a direct mode. Page 22 1229825 V. Description of the invention (19) The above steps may include the following steps: if the giant area of the B image and the common location giant block of the reference image of list 1 is ^, „a ,, ... .... & Sentence one frame mode in addition to the shed mode 'and 1 isi / soil, Ding called% < Liu Ίer 1 e block is a formula' then calculate the direct mode movement of each field i in a B frame The vector # 拉 MVB, i is as follows: f, 1 and

MVF, i =TDB, i x MV / TDDMVF, i = TDB, i x MV / TDD

MVB, i = (TDB, i -TDD) x MV / TDD 或 z = TDB,i X 256/TDDMVF,i = (Z x MV + 128) >> W = Z - 256MVB, i = (W X MV + 128) >> 8 其中，TDB，i表示一現行6欄位與一Hst 〇基準 時間距離，若此距離係從6欄位測量時，則 a (定+)一v(:距 間的時間距離2==1基準巾貞與llst 0基準巾貞之 豆浐宏一 1： 咕攸1St 1基準幀測量時，則對 ::式二步驟可包括以下步驟：若6圖像之巨區塊為-式，乂1 i st ί 基準圖像之共位置巨區塊為-攔位模 下等式計算—β土巾貞圖1在日^間上為Β圖像之後時，則從以 用一 llSt !基準鴨之直^=;動向量，與謂，其中使 算直接模式運動向：攔位°内共位置區塊之運動資訊，計 MVF =TDB X Mv〇 /TDD>〇 第23頁 1229825 五、發明說明（20) MVB — (TDB -TDD， 〇) X MVO / TDD， 〇 Z = TDB X 256 / TDD,〇MVF= (Z x MVO + 128) » « W = Z - 256MVB = (W χ MVO + 128) » 8 :“ T^B表不j見行β幀與一 ! i s t 〇基準幀之間的 離，右此距離係巾貞測量時，'則對其指定一正（+ )號，】， 此距離係從丨ist 〇基準幀測量時，則對其指定一。右 號；TDD，〇表示Hst !基準幢之攔位〇與一 Hst 〇基準 之間的時間距離，若此距離係從丨丨 二時，則對其指定一正⑴號，若此距離 =測量時’則對其指定一負㈠號；⑽。表示_直土準 式用Ust 1基準幀之攔位〇内共位置區塊之運動向 果 =以上步驟可包括以下步驟：若B圖像之巨區塊為 令貞杈式’而list 1基準圖像之共位置巨區塊 允為二 式，且11 s t 1基準圖像在時間上為B 二 *立杈 下：式計算一“貞之直接模式運動向量mvf與:二’ 算直接模式運動向量：，、位置£塊之運動貧訊，計 MVF =TDB χ MV1 / TDD, 1 MVB = (TDB -TDD， 1) x MV1 / TDD， 1 或 z = TDB X 256 / TDD，1MVF= (Z x MV1 + 128) >> s w = Z - 256MVB = (W X MV1 + 128) >〉 8 其中，TDB表示一現行“貞與一 Hst 〇基準幢之間的時間距 第24頁 1229825 五、發明說明（21) 離，若此距離係β &、αϊ胃 > 此距離係從list:/、隹測置時，則對其指'一正⑴號，若 號；⑽，】表示工丄基i準:二量時’則;其，1 一負㈠ 之間的日# η π私 基準幀之攔位1與一 st 〇基準欄位 量時，則i复於二若此距離係從list 1基準鴨之搁位1測 攔位測量4二=，，若此距離=1 ls"基準 、對/、才曰疋一負（-）號；而MV1表示一直接槿 1 s t 1基準巾貞之攔位1内共位置區塊之運動向量。' =才本發明另一層面，其中提供一種方法，用以計算動畫 飧碼糸統:B(雙預測）圖像之直接模式運動向量以取出b ^ ^之直接杈式運動向量。此方法包括以下步驟：若一直接 二式用11 = 1基準圖像内共位置巨區塊為一，在内，模式 日y則以空間冗餘度為基礎，從B圖像之欲編碼巨區塊的 預測並計算1is"基準圖像與list 1基準圖像及運 ，向，’並計算B圖像之直接模式運動向量。MVB, i = (TDB, i -TDD) x MV / TDD or z = TDB, i X 256 / TDDMVF, i = (Z x MV + 128) > > W = Z-256MVB, i = (WX MV + 128) > > 8 where TDB, i represents the distance between a current 6 field and an Hst 〇 reference time, if this distance is measured from the 6 field, then a (definite +) a v (: distance between Temporal distance 2 == 1 reference frame and llst 0 reference frame and doujin macro 1: 1: Guyou 1St 1 When measuring the reference frame, the second step of the formula can include the following steps: If the 6 image is a huge block It is-, and the co-located giant block of the base image is -1 i st 计算 The calculation of the block equation below-β soil towel Figure 1 is after the B image on the day, then it is used A llSt! Reference Duck Straight ^ =; motion vector, and predicate, in which the direct mode motion is calculated: the motion information of the co-located blocks within the block °, calculated MVF = TDB X Mv〇 / TDD> 〇 page 23 1229825 V. Description of the invention (20) MVB — (TDB -TDD, 〇) X MVO / TDD, 〇Z = TDB X 256 / TDD, 〇MVF = (Z x MVO + 128) »« W = Z-256MVB = ( W χ MVO + 128) »8:" T ^ B represents the distance between the β frame and the one! Ist reference frame. When this distance is measured, you will assign a positive (+) sign to it,], and this distance will be assigned one when measured from the reference frame. Right number; TDD, 〇 means Hst ! The time distance between the reference block 〇 and a Hst 〇 reference. If this distance is from 丨 丨 2nd, it will be assigned a positive sign, if this distance = during the measurement, it will be assigned a negative ㈠ 号; ⑽. Representation _ Straight-ground quasi-type uses the Ust 1 reference frame's position 0. The movement of the co-located blocks within the position = the above steps can include the following steps: If the giant block of the B image is a stance 'And the co-located giant block of the list 1 reference image is allowed to be of the second type, and the 11 st 1 reference image is B 2 in time * standing position: the formula calculates a "Zhen Zhi direct mode motion vector mvf and: two' Calculate the motion vector in direct mode: ,, motion loss at the position £ block, calculate MVF = TDB χ MV1 / TDD, 1 MVB = (TDB -TDD, 1) x MV1 / TDD, 1 or z = TDB X 256 / TDD, 1MVF = (Z x MV1 + 128) > > sw = Z-256MVB = (WX MV1 + 128) >> 8 where TDB stands for a current "Zheng and a Hst 〇 benchmark building Time interval on page 24 1229825 V. Description of the invention (21) If this distance is β &, αϊweit > When this distance is measured from list: /, then it means '一 正 ⑴ 号' ， 若 号 ； ⑽ ，】 represents the basic standard of work: two times when you're; then, the day between 1 and negative # # η π private reference frame block 1 and a st 〇 reference field amount, Then i is more than two. If this distance is measured from the base 1 of the reference duck on the list 1 and the measurement of the second stop 4 ==, if this distance = 1 ls " the reference, right, and only a negative (-) sign; and MV1 represents the motion vector of a co-located block in Block 1 of a direct frame 1 st 1 reference frame. '= Another aspect of the present invention, in which a method is provided for calculating an animation code system: a direct mode motion vector of a B (bi-prediction) image to extract a direct branch motion vector of b ^. This method includes the following steps: if a direct binary equation uses 11 = 1 co-located giant blocks in the reference image as one, the pattern day y is based on the spatial redundancy, and the giant image is encoded from the desire of the B image. Block prediction and calculation of 1is " reference image and list 1 reference image and operation, and 'and calculate the direct mode motion vector of B image.

Ua疋b以ΛΛ驟可包括以下步,驟：若欲編碼巨區塊之鄰 之美m ® # 表不不同的基準圖像時，則選取索引最小 基準圖像’作為每一list之基準圖像。 Π，：上步驟可包括以下步驟：若欲編碼巨區塊之鄰區 該基準“示索引相同之基準圖像時，則選取 备旱圖像作為母一 list之基準圖像。 此外，以上步驟可包括以下步驟：若^ ^ ^ ^ ^ ^ ^ ^ ^ ^ ，有任—個為，在内，模式時 二1 = 為0 ;選取一運動向量，其方向須與 母hst之基準圖像從一鄰區塊起算之時間位置的方向相Ua 疋 b can include the following steps in ΛΛ steps: If you want to encode the beauty of the neighbors of the giant block m ® # is a different reference image, then select the smallest index reference image 'as the reference image for each list . Π: The previous step may include the following steps: if a reference image with the same index as the reference image of a neighboring block of a giant block is to be encoded, a drought-prepared image is selected as the reference image of the parent list. In addition, the above steps It can include the following steps: If ^ ^ ^ ^ ^ ^ ^ ^ ^ ^, there is any one of them. In the mode, 1 = 0; choose a motion vector whose direction must be from the reference image of the parent hst. Direction of the time position from a neighboring block

1229825 五、發明說明（22) ^及，經由中值運算取得每— 只選取二個向量1:個；動向量時，則從該ί塊 運算，取得每-ust之；=括所選取運動向量之中值 :外W上步驟可包括以下步驟：若益、 — 式之有效基準圖像索引時，則將u 其母一list模 list 1基準圖像索引 基準圖像索弓I與 向量設定為〇。 ......字母一llst模式之運動 藉由以下參照附圖… 述及其他目的、特徵與優Ϊ 更加清晰了解本發明上 [實施方式] 本發明建議一種方法，可於一 :共位至巨區塊為一，在内，模式時二以二：準圖像 動向量；本發明亦建議一種方法，可式運 時間上為-B圖像之後的情況、1基準圖像在 為-β圖像之前的情況中，獲得直接模以：r間上 Πί:建種計算直接模式運動向量的方1 :運指定-符號，簡化直接模 與基Ϊ圖:：：Ϊ:而可不考慮直接模式用 5 3以Β圖由後於，在一圖像層上切換-幀模式與-攔位模 ^因此1 St 1基準圖像可編碼成幅模式或欄位 掩；圖像之巨區塊與1iStl基準圖像之共位置 巨區塊有四種幀/欄位編碼組合。 第26頁 1229825 五'發明說明（23) —---------- [Uhst 1基準圖像之妓 況： ，、位置巨£塊為，在内，模式之情 如圖3f與4f所干，直^ 、 巨區塊可為，在不，y妾模式用1 ist 1基準圖像内的共位置 因為此模式的内二，，不論基準圖像之時間位置如何。 接模式運動向ς2有運動資訊，所以習用方法只將直 後解碼的圖俊 並疋義一 list 0基準圖像為最 M 1豕。然而，怒田女、+ ^ 此，本發明以空間冗 :-法保證高編碼效率。因 塊的鄰區塊，預測虚外^為基礎，從B圖像之欲、編碼巨區 向量。 十异llst 〇與list 1基準圖像及運動 示考慮一二1 ist模式之基準圖像索引。圖5顯 工日】几餘度而使用鄰區塊A、b、c運動向量作 3量預測。 P； _ 區塊E之運動向量預測。一，，…π扭D、U逆π丨口）里丨「 〜〜若鄰區塊A、Β、f古τ η 其中最小的一伽宜唯有同的基準圖像索引時，係決定以 引。 、 土 >圖像索引作為直接模式用基準圖像索 以此索5 I ΐ 一個具有相同的基準圖像索引時，係決定 〜=為直接模式用基準圖像索引。 索引作為H ί有相同的基準圖像索引0夺，係決定以此 此Γ 5 式用基準圖像索弓卜 獐彳曰每11 s 11杈式之運動向量是經由以下運動向量預剎 ς H此時，若鄰區塊A、B、C中，有任一個為，在内， 。時，該-區塊之list 〇與Ust丨運動向量都設定為1229825 V. Description of the invention (22) ^ and obtain each through median operation-only select two vectors 1: motion vectors; when moving vectors, from this til block operation to obtain each -ust; = include the selected motion vector Median: The above steps may include the following steps: If the effective reference image index of the formula is, set u to its parent list list list 1 reference image index reference image cable I and vector as 〇. ...... The movement of the letter-llst mode is described below with reference to the drawings ... A description of other purposes, features, and advantages A clearer understanding of the present invention [Embodiment] The present invention proposes a method that can be used in one: The giant block is one. In the mode, two and two are used in the mode: quasi-image motion vector; the present invention also proposes a method that can be used after the -B image in time, and the reference image is In the case before the β image, the direct module is obtained by: r 上 上 Πί: the formula for calculating the direct mode motion vector 1: operation designation-symbol, simplifying the direct module and the base map: ::: Ϊ: without considering direct The mode uses 5 3 in the B picture from behind, switch on an image layer-frame mode and-blocking mode ^ So 1 St 1 reference image can be encoded into a frame mode or a field mask; the huge block of the image There are four types of frame / field coding combinations for the co-located giant block with the 1iStl reference image. Page 26 1229825 Five 'invention description (23) ----------- [Prostitutes of Uhst 1 reference image:,, location is huge, block is included, the mode is shown in Figure 3f and For 4f, the straight and giant blocks can be, in no, y 妾 mode uses the common position in the 1 ist 1 reference image because of the inner 2 of this mode, regardless of the time position of the reference image. The connection mode motion direction 2 has motion information, so the conventional method only decodes the picture that is decoded directly and defines a list 0 reference image as the most M 1 豕. However, Nu Tianyu, + ^ Therefore, the present invention ensures high coding efficiency with a spatial redundancy:-method. Based on the neighboring blocks of the block, the prediction of the imaginary outer space is used as the basis for encoding the huge region vector from the desire of the B image. Ten different llst 〇 and list 1 reference image and motion show the reference image index considering the 12 1 ist mode. Fig. 5 shows the working day] using the neighboring block A, b, and c motion vectors for three-quantity prediction. P; _ motion vector prediction of block E. First,… π twist D, U inverse π mouth) ～ ～~ If the neighboring block A, B, f τ η η is the smallest one, the same reference image index should be used. The index is used as the reference image for direct mode. 5 I ΐ When one has the same reference image index, it is determined that ~ = is the reference image index for direct mode. The index is H ί With the same reference image index of 0, it is decided to use the reference image in this Γ 5 formula. The motion vector of each 11 s 11 branches is pre-brought through the following motion vectors. At this time, if Among the neighboring blocks A, B, and C, any of them is, and the time, the list 〇 and Ust 丨 motion vectors of the -block are both set to

第27頁 1229825Page 27 1229825

五、發明說明（24) k 一鄰區塊選取一運細&旦 1 lSt Μ ^ - ΛΑ « 運動向^ ’其方向須與上述為每一 值運算取得每-lis"莫式之運動向ί方向相同，並經由中 鬼：有二個方向相同的ί動向量時，只選取該 &塊中的二個運動向吾 k取口％ m 動向里之一，並包括在中值運算中。 1美準m禮’二丨疋主'、法從一鄰區塊導出有效的1 ist 0與1 ist 之運動向量設定為。。 又疋為〇 ’並將母-Ust模式 =]直接模式用Ust !基準圖像在時間上為β圖像之後的情 情況1 : Β圖像之巨區塊盥j i t ]其 都為巾刚 充」…丨基準圖像之共位置巨區塊 =圖3U3h所示，nst !基準圖像内的共位置巨區塊可有 一或二個運動向量。在本發明中，若該共位置區塊有二個 運動向篁]則選取其中之一（L〇 MV或Ll MV)，並從選取的 運動向量導出直接模式運動向量（以下會以選取[〇 MV[LIST 0運動向量]的情況為基礎，加以說明）。 _ 因此，圖3a、3c ;圖3b、3d、3e ;圖3g、3h可分別簡單地 表示為圖6a ;圖6c ;及圖6b。 若直接模式用list 0基準圖像與iist 1基準圖像在時間上 刀別為位於B圖像之如與之後（圖6 a )，或如直接模式用 list 0基準圖像與list 1基準圖像在時間上都為位於b圖 像之後’且1 i s t 0基準圖像在時間上為1 i s t 1基準圖像之 後（圖6 b)’則計异直接模式運動向量Μ V F與Μ V B如下：V. Description of the invention (24) k Adjacent block selects a transport detail & Dan 1 lSt Μ ^-ΛΑ «Motion direction ^ 'The direction must be the same as above for each value operation to obtain each -lis " Mo style motion direction ί in the same direction and passing through the ghost: when there are two ί motion vectors in the same direction, only one of the two motions in the & block is taken to move to% k and the motion is included in the median operation . 1 US standard m ceremony 'two 丨 lord', method to derive effective 1 ist 0 and 1 ist motion vector from a neighboring block is set to. . Again, it is 0 ′ and the mother-Ust mode =] The direct mode uses Ust! The reference image is temporally after the β image. Case 1: The large block of the Β image is jit … 丨 Co-location macroblocks in the reference image = As shown in Figure 3U3h, the co-location macroblocks in the nst! Reference image can have one or two motion vectors. In the present invention, if the co-location block has two motion directions 篁], one of them (L0MV or L1 MV) is selected, and a direct mode motion vector is derived from the selected motion vector. Based on the case of MV [LIST 0 motion vector], it will be explained). _ Therefore, Figures 3a, 3c; Figures 3b, 3d, 3e; Figures 3g, 3h can be simply expressed as Figure 6a; Figure 6c; and Figure 6b. If the direct mode uses the list 0 reference image and the iist 1 reference image in time, it is located after the B image (Figure 6a), or if the direct mode uses the list 0 reference image and the list 1 reference image. If the image is temporally located after the b image 'and the 1 ist 0 reference image is temporally 1 ist 1 after the reference image (Figure 6 b)', then the direct mode motion vectors MV VF and MV VB are calculated as follows:

第28頁 1229825 五、發明說明（25)Page 28 1229825 V. Description of the invention (25)

MVF = TDB X MV / TDDMVF = TDB X MV / TDD

MVB = (TDB - TDD) χ MV / TDD 其中TDB表示現行B幀與~ i丨s t 〇基準幀之間的時間距 離T D D表示1丨s t 1基準幀與1 i s t 〇基準幀之間的時間 距離。 』 為方便之故，在直接模式運動向量MVF與MVB之計算上應 一位元運算，以上等式可表示如下： 一 Z = TDB χ 256 / TDDMVF - (z x MV + 128) >> 8 W = Ζ - 256MVB = (W χ MV + 128) >> 8 右直接核式用list 0基準圖像與Hst i基準 都為位於B圖像之後，且list 〇基準圖像在 、0 1基準圖像之前(圖6c) ’則計算直接模式：Π lst MVB如下： 沒動向篁MVF與MVB = (TDB-TDD) χ MV / TDD where TDB represents the time distance between the current B frame and the ~ i 丨 s t 〇 reference frame T D D represents the time distance between the 1 s t 1 reference frame and the 1 i s t 〇 reference frame. 』For the sake of convenience, the calculation of the direct mode motion vectors MVF and MVB should be a one-bit operation. The above equation can be expressed as follows:-Z = TDB χ 256 / TDDMVF-(zx MV + 128) > > 8 W = ZO-256MVB = (W χ MV + 128) > > 8 The right direct kernel list 0 reference image and Hst i reference are both behind the B image, and the list 〇 reference image is at 0 0 Before the reference image (Figure 6c), the direct mode is calculated: Π lst MVB is as follows:

MVF = — TDB χ MV / TDDMVF = — TDB χ MV / TDD

MVB = — (TDB + TDD) χ MV / TDD 此一等式可表示如下：MVB = — (TDB + TDD) χ MV / TDD This equation can be expressed as follows:

2 = - TDB χ 256 / TDDMVF = (Z x MV + l2R W = Z 一 256MVB = (W x MV + 128) >> 8 ”>8 區塊 螓 情況2 ·· B圖像之巨區塊與i丨s t i基準圖像 都為攔位模式 ，、位置巨 圖7a至7d顯示B圖像之巨區塊與Hst 1基準圖 巨區塊都為攔位模式的各種情況。B圖像之巨之共位置 運動向量，係從同位（the same parity) 品塊的每一 像内共位置區塊導出的。 1 s t 1基準圖2 =-TDB χ 256 / TDDMVF = (Z x MV + l2R W = Z-256MVB = (W x MV + 128) > > 8 '' > 8 Block 螓 Case 2 ·· B huge area of B image Both the block and i 丨 sti reference image are in stop mode, and the location giants 7a to 7d show the huge block of image B and the Hst 1 reference map giant block are in stop mode. B image The giant co-location motion vector is derived from the co-location block within each image of the same parity block. 1 st 1 reference map

1229825 五、發明說明（26) 若直接模式用list 0基準圖像與 為分別位於B圖像之前與之後（圖7a )， 土 ’圖像在時間上 Ust 0基準圖像與list !基準圖像在時5間如上直接模式用 像之後，且list 0基準圖像在時間上 都為位於B圖 腳b)，則計算-B㈣—攔位二二!，準圖像之 list 1運動向量MVF，i與MVB，i如下（卜n矣_ t式1St 〇與 i = 1表示第二攔位）： 下（卜0表不第―攔位， MVF, i = TDB, i X MVi / TDD, i MVB， i = (TDB， i - TDD, i) χ MVi / TDD i 其中，MVi表示一list 1基準幀内一攔位 運動向量，T D B，i表示一現行B攔位與_丨i s t /〇 = f區塊的 之時間距離，TDD，i表示一llst !基準攔位與上：位間 欄位間之時間距離。 〇基準 以上等式可表示如下： z = TDB，i X 256/TDD，IMVF，i = (z x MVi + κ W = Z - 256MVB, i = (W X MVi + 128) >> 8 8 如果，由於list 1基準幀内攔位i之共位置區塊有一 向量所指向的攔位係在一時間上為B圖像之後的幀 = 以直接模式用1 i s t 0與1 i s t 1基準圖像在時間上 像之後，且Hst 0基準圖像在時間上為llst ! ^圖 前（圖7c、7d)，則計算直接模Slist 〇與1131 i MVF, i與MVB，i如下： 連動向ϊ MVF, i = - TDB, i X MVi / TDD， i MVB， i = - (丁DB, i + TDD， i) χ MVi / TDD， i 128) >> 1229825 五、發明說明（27) 以上等式可表示如下·· Z = -TDB, i X 256/TDD, iMVF i ^ ί7 W - Ζ - 256MVB,i - (? x MVi + ^ MVi 基準圖像之 情況3 ·· β圖像之巨區塊為攔位模〉 共位置巨區塊為幀模式 、…而llst 圖8a至8c顯示β圖像之巨區塊為 圖像之共位置巨區塊為㈣式時的各種況而=1基準 仃巨區塊之縱座標為ycurrent， 八中，讓現 巨區塊之縱座標為Yco_1〇cated， "準圖像之共位置 係為 YC〇-i〇eated=2xYeurrent m 間建立的關 進捫办八如rt, ! · Λ t 此外，1 1 S t 0與1 i s t 1基 卓攔位刀別與llst 〇mist !基準㈣同位。 若直接模式用1ist G與1 ist 1基準圖像在時間上為分別位 於B圖像之前與之後（圖8a)，或如直接模式用Hst 〇與 1 St 1基準圖像在時間上都為位於B圖像之後，且1 i s t 〇 基準圖^象在時間上為iist i基準圖像之後（圖8b)，則計算 B幀每一欄位】之直接模式Ust 〇與iist i運動向量MVF， i與MVB，i如下：1229825 V. Description of the invention (26) If the direct mode uses the list 0 reference image and is located before and after the B image (Figure 7a), the soil image is temporally Ust 0 reference image and list! Reference image After using the image in the direct mode as above, and the reference image of list 0 is located at the bottom of figure B in time, then -B㈣—blocking two two !, the list 1 motion vector MVF of the quasi-image, i and MVB, i are as follows (bu n 矣 _t formula 1St 〇 and i = 1 represents the second stop): (bu 0 represents the first stop-MVF, i = TDB, i X MVi / TDD, i MVB, i = (TDB, i-TDD, i) χ MVi / TDD i where MVi represents a block motion vector in a list 1 reference frame, TDB, i represents an existing B block and _ 丨 ist / 〇 = The time distance of block f, TDD, i represents an llst! The time distance between the reference stop and the upper: the time between the fields. 〇 The above equation can be expressed as follows: z = TDB, i X 256 / TDD, IMVF , I = (zx MVi + κ W = Z-256MVB, i = (WX MVi + 128) > > 8 8 If, because of the block i in the list 1 reference frame, there is a block pointed to by a vector Position The frame after the B image in one time = after the temporal image with 1 ist 0 and 1 ist 1 reference images in direct mode, and the Hst 0 reference image is llst in time! ^ Before the figure (Figure 7c, 7d), then calculate the direct module Slist 〇 and 1131 i MVF, i and MVB, i is as follows: Linking direction ϊ MVF, i =-TDB, i X MVi / TDD, i MVB, i =-(ding DB, i + TDD , I) χ MVi / TDD, i 128) > > 1229825 V. Description of the invention (27) The above equation can be expressed as follows: Z = -TDB, i X 256 / TDD, iMVF i ^ 7 W-ZZ- 256MVB, i-(? X MVi + ^ MVi reference image case 3 · The giant block of the β image is a blocking mode> The co-located giant block is a frame mode, and llst Figures 8a to 8c show the β chart The image of the giant block is the common position of the image when the giant block is in the normal mode, and = 1. The vertical coordinate of the reference giant block is ycurrent. Eighth, let the vertical coordinate of the current giant block be Yco_1〇cated. The common position of " quasi-images is the relationship between YC〇-i〇eated = 2xYeurrent m established rt,! · Λ t In addition, 1 1 S t 0 and 1 ist 1 Gizo stop The knife is in the same position as llst 〇mist! If the direct mode uses 1ist G and 1 ist 1 reference images to be located before and after the B image in time (Figure 8a), or if the direct mode uses Hst 〇 and 1 St 1 reference images to be located in time After the B image, and the 1 ist 〇 reference image is temporally after the iist i reference image (Figure 8b), the direct mode Ust 〇 and the iist i motion vector MVF, i With MVB, i is as follows:

MVF, i =TDB, i x MV / TDDMVF, i = TDB, i x MV / TDD

MVB, i = (TDB, i -TDD) x MV / TDD 以上等式可表示如下： Ζ ~ TDB, ι χ 256 / TDDMVF,i = (Ζ χ MV + 128) >> 8 W = Ζ - 256MVB, i = (W χ MV + 128) >> 8 如果，由於list 1基準幀内共位置區塊有一運動向量所指 向的幀係在時間上為B圖像之後，所以直接模式用丨丨st 〇 第31頁 1229825 五、發明說明（28) --- 與11st 1基準圖像在時間上都位於B圖像之後，且丨ist 〇 基準圖—像在時間上為1 i s t 1基準圖像之前（圖8 c)，則計算 一 B幀每一攔位i之直接模式list 0與list 1運動向tMVIT i與MVB，i如下： ’MVB, i = (TDB, i -TDD) x MV / TDD The above equation can be expressed as follows: ZO ~ TDB, ι χ 256 / TDDMVF, i = (Z χ MV + 128) > > 8 W = ZZ- 256MVB, i = (W χ MV + 128) > > 8 If the frame pointed to by a motion vector in the co-location block in the reference frame of list 1 is behind the B image in time, direct mode is used.丨 st 〇 page 31 1229825 V. Description of the invention (28) --- The 11st 1 reference image is located behind the B image in time, and the ist 〇 reference image—the image is 1 ist 1 reference image in time Like before (Figure 8c), calculate the direct mode list 0 and list 1 of each stop i of a frame B to tMVIT i and MVB, i is as follows: '

MVF,1 = -TDB, i X MV / TDDMVF, 1 = -TDB, i X MV / TDD

MVB,1 = - (TDB, i +TDD) x MV / TDD 以上專式可表示如下： Z TDB, i χ 256/TDDMVF, i = (Z x MV + 128) >> 8 W = Z - 256MVB, i = (W x MV + 128) >> 8 其中，TDB，i表示現行^闌位與Ust 〇基準攔位之間的 距離，TDD表示Hst }基準幀與丨丨以〇基準幀之間的時間 距離，MV表示直接模式用list 1基準幀内共位置區塊之 而list 1基準圖像之共 情況4 ·· B圖像之巨區塊為幀模式 位置巨區塊為攔位模式 圖9a至9c顯示B圖像之巨區塊為幀模式，而Ust i其 像之共位置巨區塊為欄位模式時的各種情況。豆 wMVB, 1 =-(TDB, i + TDD) x MV / TDD can be expressed as follows: Z TDB, i χ 256 / TDDMVF, i = (Z x MV + 128) > > 8 W = Z- 256MVB, i = (W x MV + 128) > > 8 where TDB, i represents the distance between the current ^ stop and the Ust 〇 reference stop, and TDD means Hst} reference frame and 丨 丨 reference frame The time distance between them, the MV indicates that the direct mode uses the common position blocks in the list 1 reference frame and the common situation of the list 1 reference image 4 ·· The giant block of the B image is the frame mode and the giant block is the stop. Modes Figures 9a to 9c show various situations when the giant block of image B is in frame mode, and the common location giant block of Ust i is in field mode. Beans w

行巨區塊之縱座標為Ycurrent，Ust」基準圖像 K 巨區塊之縱座標為Yco-located，且此二座標間建立的置 係為Yc〇-l〇cated = Ycurrent/2。此外，由於Ust 丨美 之欄位。比其欄位1在時間上較為靠近B圖像，所以使土 ' 位0之共位置區塊的運動資訊計算直接模式運動向 若直接模式用list (^list i基準圖像在時 於B圖像之前與之後（圖9a)，或如直接模式用⑴^刀與別位The vertical coordinate of the row giant block is Ycurrent, Ust. The benchmark image of the K giant block is Yco-located, and the relationship established between these two coordinates is Yc0-located = Ycurrent / 2. In addition, due to the Ust 丨 beauty column. It is closer to the B image in time than its field 1, so the motion information of the co-located block of the soil bit 0 is calculated in direct mode. If the direct mode uses list (^ list i Like before and after (Figure 9a), or as direct mode, use ⑴ ^ 刀 and other positions

1229825 五、發明說明（29) 1 1 S t 1基準圖像在時間上都位於B圖像之後，且丨丨$七〇美 準圖像在時^上為list i基準圖像之後（圖9b)，則計算^ B幀之直接杈式ilst 〇與list 1運動向量Μνρ與mvb如下： MVF =TDB X MVO / TDD，0 、 MVB = (TDB -TDD，〇) x MVO / TDD，〇 以上等式可表示如下： Z = TDB x 2 5 6 / TDD，〇MVF= (Z x MVO + 128) >> 8 W = Z _ 256MVB = (W x MVO + 128) >> 8 如果，由於list 1基準幀欄位〇之共位置區塊有一運動向 量所指向的幀欄位在時間上為B圖像之後，所以直接模式 用11 s t 0與11 s t 1基準圖像在時間上都位於B圖像之後， 且list 0基準圖像在時間上為list i基準圖像之前（圖 9c)則计异直接模式list 0與list 1運動向量MVF與MVB 如下： MVF = -TDB X MVO / TDD,0 MVB = - (TDB+TDD，0) X MVO / TDD，〇 以上等式可表示如下： 128) >>1229825 V. Description of the invention (29) The 1 1 S t 1 reference images are located behind the B image in time, and the $ 70 US standard image is behind the list i reference image at time ^ (Figure 9b ), Calculate the direct frame ilst 〇 and list 1 motion vectors Μνρ and mvb of ^ B frames as follows: MVF = TDB X MVO / TDD, 0, MVB = (TDB-TDD, 〇) x MVO / TDD, 〇 or more The formula can be expressed as follows: Z = TDB x 2 5 6 / TDD, 〇MVF = (Z x MVO + 128) > > 8 W = Z _ 256MVB = (W x MVO + 128) > > 8 If, Since the common position block of the reference frame field 0 of list 1 has a frame field pointed to by the motion vector in time, it is the B image in time, so the direct mode uses 11 st 0 and 11 st 1 reference images to be located in time. After the B image and the list 0 reference image is temporally before the list i reference image (Figure 9c), the direct mode list 0 and list 1 motion vectors MVF and MVB are as follows: MVF = -TDB X MVO / TDD , 0 MVB =-(TDB + TDD, 0) X MVO / TDD, above equation can be expressed as follows: 128) > >

Z = - TDB X 256/TDD,0MVF= (Z χ MVO W = Z - 256MVB = (W x MVO + 128) >> 8 其中，TDB表示現行B巾貞與list 〇基準巾貞之間的時間距離， TDD，〇表示list ]基準巾貞之攔位〇與list 〇基準攔位 間距離，^〇表示直接模式用Ust 0與11討 位置區塊之運動向量。 i Π /、 [3]直接模式用1ist 1基準圖像在時間上為Β圖像之前的情 1229825Z =-TDB X 256 / TDD, 0MVF = (Z χ MVO W = Z-256MVB = (W x MVO + 128) > > 8 where TDB represents the time between the current B frame and the list 〇 benchmark frame Distance, TDD, 〇 indicates list] The distance between the reference frame of reference frame 0 and list 〇 the reference frame, ^ 〇 represents the direct vector using the Ust 0 and 11 to determine the motion vector of the location block. I / / [3] direct mode Using 1ist 1 reference image is temporally before the B image 1229825

位 在此情況時，list 〇*list i 於B圖像之前。 j上都總是 區塊 ，況i 1圖像之巨區塊mist 圖像 都為幀模式 立置巨 如圖4a至4h所不，1 lst J基準圖像内共位置區 運動向量。在本發明中，若該共位置區塊有：：運: 向，:則選取其中之一α〇 Mv或以ΜΌ，並從選取的索 向里導出直接模式運動向量（以下會以選取[〇 Μν[ 1 i Μ ， 運動向量]的情況為基礎，加以說明）。 因此，圖4a、4c、4e、4g、4h及圖4b、4d可分別簡單地表 不為圖1 Oa及圖1 Ob。 若直接模式用1 ist 0基準圖像在時間上為直接模式用丨丨st 基準圖像之前，則計算直接模式運動向量Μνρ與MVB如In this case, list 〇 * list i precedes the B picture. All blocks on j are always blocks. In addition, the giant block mist image of the i 1 image is in frame mode. The vertical giant is shown in Fig. 4a to 4h. The motion vector of the co-location area in the 1 lst J reference image. In the present invention, if the co-location block includes :: direction,:, select one of αomv or Mm, and derive the direct mode motion vector from the selected cable direction (hereinafter, select [〇 (Based on the case of Μν [1 i Μ, motion vector]). Therefore, Figs. 4a, 4c, 4e, 4g, 4h and 4b, 4d can be simply expressed as Fig. 10a and Fig. 1b, respectively. If the direct mode use 1 ist 0 reference image is temporally direct mode use before the st reference image, calculate the direct mode motion vectors Μνρ and MVB as

mVF = TDB X MV / TDDmVF = TDB X MV / TDD

MVB = (TDB - TDD) x MV / TDD 其中，TDB表示現行B幀與一list 〇基準幀之間的時間距MVB = (TDB-TDD) x MV / TDD where TDB is the time interval between the current B frame and a list 〇 reference frame

離，TDD表示一list !基準巾貞與list 〇基準巾貞之間的時間 距離，Μ V表不直接模式用丨丨s t i基準圖像内共位置區塊的 運動向量。 以上等式可表示如下： TDB X 256 / TDDMVF = (Z x MV + 128) >> 8 Z - 256MVB = (W X MV + 128) >> 〇 W =TDD represents the time distance between a list! Reference frame and a list 〇 reference frame. The MV table uses the motion vector of co-located blocks in the reference image in the direct mode. The above equation can be expressed as follows: TDB X 256 / TDDMVF = (Z x MV + 128) > > 8 Z-256MVB = (W X MV + 128) > > 〇 W =

1229825 五、發明說明（31) 若1 ist 0基準圖像在時間上為1 ist 1基準圖像之後，則計 算直接模式運動向量MVF與MVB如下（圖10b):1229825 V. Description of the invention (31) If the 1 ist 0 reference image is 1 ist 1 reference image in time, the direct mode motion vectors MVF and MVB are calculated as follows (Figure 10b):

MVF = - TDB X MV / TDDMVF =-TDB X MV / TDD

MVB = - (TDB + TDD) x MV / TDD 此等式可表示如下·· z = - TDB X 256 / TDDMVF = (Z X MV + 128) >> 8 w = Z - 256MVB = (W x MV + 128) >> 8 其中，TDB表示現行B幀與一list 0基準幀之間的時間距 離’ T D D表示1 i s t 1基準幀與1 i s t 〇基準幀之間的時間距 離，MV表示直接模式用1 ist 1基準圖像内共位置區塊的運 動向量。 清況2 :B圖像之巨區塊與nst 1基準圖像之共位置巨區塊 都為攔位模式 若直接模式用list 0基準圖像在時間上為直揍模式用Hst 1基準圖像之前，則計算一 β幀每一攔位i之直接模式1 i s t 0與list 1運動向量MVF，i與MVB，i如下（圖iia、ilb): i = TDB, i x MVi / TDD, i MVB, i = (TDB, i - TDD, i) x MVi / TDD, i 以上等式可表示如下： i >>MVB =-(TDB + TDD) x MV / TDD This equation can be expressed as follows: z =-TDB X 256 / TDDMVF = (ZX MV + 128) > > 8 w = Z-256MVB = (W x MV + 128) > > 8 where TDB represents the time distance between the current B frame and a list 0 reference frame 'TDD represents the time distance between 1 ist 1 reference frame and 1 ist 〇 reference frame, and MV indicates direct mode Use the motion vectors of the co-located blocks in the 1 ist 1 reference image. Case 2: Both the giant block of the B image and the co-located giant block of the nst 1 reference image are in blocking mode. If the direct mode uses list 0, the reference image is temporally in direct mode, and the Hst 1 reference image is used. Before, calculate the direct mode 1 ist 0 and list 1 motion vectors MVF, i and MVB of each stop i of a β frame, i is as follows (Figure iia, ilb): i = TDB, ix MVi / TDD, i MVB, i = (TDB, i-TDD, i) x MVi / TDD, i The above equation can be expressed as follows: i > >

TDB，i χ 256/TDD，iMVF，i = “ χ 評丄 复〜z - 256MVB，i = (f x MVi + 128) >> gTDB, i χ 256 / TDD, iMVF, i = "χ Comment 〜 z-256MVB, i = (f x MVi + 128) > > g

ί l ^ ^ -List 〇 ^ ^ ^ ^ ^ fE 位之門的性一 基準攔位與1ist 0基準 曰夺間距離，MVi表示直接模式用list i基準攔 1229825 五、發明說明（32) 内共位置區塊之運動向量。 如果，由於list 1基準幀内欄位土之共位置區塊有一運動 向量指向一時間上為後之幀内的攔位，所以〇基準圖 像在時間上為list】基準圖像之前，則計算直接模式Ust 〇與list 1運動向量MVF，i與MVB’i如下（圖llc、lld): MVF, i = ~ TDB, i χ MVi / TDD, i MVB, i - - (Π)Β，i + TDD, i) x MVi / TDD，I 以上等式可表示如下·· Z -- -TDB,i X 256/TDD,lMVF,i = (Z x MVi + 128) » 8 w = Z - 256MVB, i = (W X MVi + 128) >> 8 表ΐ:現行_位與List G基準攔位之間的時 ，1示11 s t 1基準攔位與1 i s t 0基準攔位之 位置區塊之運動向量表，直接換式用llst 1基準攔位内共 之巨區塊為欄位模式，而list 共位置巨區塊為幀模式 α诼之ί l ^ ^ -List 〇 ^ ^ ^ ^ ^ fE Position of the gate is a reference stop and the 1ist 0 reference is the distance between them. MVi indicates that the direct mode uses the list i reference to stop 1229825. 5. Description of the invention (32) The motion vector of the location block. If there is a motion vector in the co-location block of the field in the reference frame of list 1 pointing to a stop in the frame that is later in time, so the reference image is list in time] before the reference image, then calculate The direct mode Ust 〇 and list 1 motion vectors MVF, i and MVB'i are as follows (Figures 11c, 11d): MVF, i = ~ TDB, i χ MVi / TDD, i MVB, i--(Π) Β, i + TDD, i) x MVi / TDD, the above equation can be expressed as follows: Z--TDB, i X 256 / TDD, lMVF, i = (Z x MVi + 128) »8 w = Z-256MVB, i = (WX MVi + 128) > > 8 Table ΐ: When the current _ bit and the List G reference stop, 1 shows the movement of the position block of 11 st 1 reference stop and 1 ist 0 reference stop Vector table, the direct use of the llst 1 reference block in the total block is the field mode, while the list common position block is the frame mode α 模式 之

ΓΛ接圖Τ之？1st °基準圖像在時間上為直接模式用llst 基準圖像之* ’則計算一“貞每 ：T 丄運動向娜，,細，i如下二 ’，1 -TDB，1 x MV / TDD MVB, i = (TDB, i -TDD) X MV / TDD 以上等式可表示如下： ;:ΤΙ心厂，=(Z X MV + 128)〉” Z ' 256MVB,i = (W x MV + 128) » 8 1229825 五、發明說明（33) 其中，TDB，i表示現行^闌位與Ust 〇基準搁位之間的 距離，TDD表示1 ist 1基準幀與1 ist 0基準巾貞之間的 u 距離，MV表示直接模式用1 i st 1基準幀内共位置區日、間 動向量。 °° A之運 如果，由於list 1基準幀内共位置區塊有一運動向旦匕 一時間上在後的幀，所以1 i st 0基準圖像在時間上二心向 1基準圖像之後，則計算一B幀每一攔位i之直接模式”1 · lst 0與list 1運動向量MVF,i與MVB，i如下（圖;[2b) : S ^Is ΓΛ connected to the figure Τ? The 1st ° reference image is in direct mode in time. * 'Of the llst reference image is used to calculate a "Zhen: T 丄 motion direction Na ,, thin, i as follows two', 1-TDB, 1 x MV / TDD MVB , i = (TDB, i -TDD) X MV / TDD The above equation can be expressed as follows:;: ΤΙ 心 厂 ， = (ZX MV + 128)〉 ”Z '256MVB, i = (W x MV + 128)» 8 1229825 V. Description of the invention (33) Among them, TDB, i represents the distance between the current ^ appendix and the Ust 〇 reference shelf, TDD represents the u distance between the 1 ist 1 reference frame and the 1 ist 0 reference frame, MV Represents the day-to-day and time-to-day motion vectors of the common location area in the 1 i st 1 reference frame for direct mode. °° If A, since the co-located block in the reference frame of list 1 has a frame moving in time, the 1 i st 0 reference image is temporally centered to the 1 reference image. Then calculate the direct mode of each block i in a B frame "1 · lst 0 and list 1 motion vectors MVF, i and MVB, i is as follows (Figure; [2b): S ^

MVF, i = ~TDB, i X MV / TDDMVF, i = ~ TDB, i X MV / TDD

MVB, i = - (TDB, i +TDD) x MV / TDD 以上等式可表示如下： z = - TDB,i x 256/TDDMVF,i = (Z x My + J2 W = Z - 256MVB, i = (W x MV + 128) >> 8 8 其中，TDB’i表示現行6攔位與Ust 〇 距離，m表示llst i基準帕mist 〇基旱準^位之之間的時間 距離，肝表示直接模式用list 1基準鴨内妓位0的時間 動向量。 +糨内共位置區塊之運 情況4 :B圖像之巨區塊為賴 位置巨區塊為攔位模式 而llst 1基準圖像之共 因為1 i s t 1基準幀之攔位 近β圖像，所以使用攔位〗n立、ί〇在時間上較靠 算直接模式運動向量。 /、位置區塊的運動資訊，計 若直接模式用1 i ς十η |、住。 1基準圖像之前，料Y 一 土為直接模式用… 攔位1之直接模式1 i s t 1229825MVB, i =-(TDB, i + TDD) x MV / TDD The above equation can be expressed as follows: z =-TDB, ix 256 / TDDMVF, i = (Z x My + J2 W = Z-256MVB, i = ( W x MV + 128) > > 8 8 where TDB'i represents the distance between the current 6 stop and Ust 〇, m represents the time distance between the llst i reference patist ist base 准, and liver indicates direct The pattern uses the time motion vector of the prostitute position 0 in the list 1 reference duck. + The situation of the common position block in the frame 4: The huge block of the B image is based on the location giant block and the llst 1 reference image. In total, because of the near β image of the 1 ist 1 reference frame, the use of the block 〖n 立, ί〇 is more dependent on the direct mode motion vector in time. /, The motion information of the location block, if the direct mode is used Use 1 i ς 十 η |, live. 1 Before the reference image, the material Y is used for the direct mode ... Stop 1's direct mode 1 ist 1229825

0與list 1運動向量MVF與MVB如下（圖133): MVF =TDB x MV1 / TDD, 1 MVB = (TDB -TDD，1) x MV1 / TDD，1 以上等式可表示如下： Z - TDB x 256 / TDD，1MVF= (Z x MVl + 128) >> 8 W = Z - 256MVB = (W x MV1 + 128) >> 8 其中，TDB表示現行B幀與Li st 0基準幀之間的時間距離， TDD, 1表示list 1基準幀之攔位1與1丨时〇基準攔位之間的 時間距離，MV1表示直接模式用list !基準幀之攔位i内共 位置區塊之運動向量。 如，，由於list 1基準幀欄位i η之共位置區塊有一運動 向量指向時間上在後之幀的攔位，所以1 i st 〇基準圖像在 時間上為1 ist 1基準圖像之後，則計算直接模式Ust 〇與 list 1運動向量MVF與MVB如下（圖13b): MVF = -TDB X MV1 / TDD, 1 MVB = - (TDB + TDD,1) x MV1 / TDD, 1 以上等式可表示如下： Z - - TDB x 25 6/TDD, 1MVF= (Z χ MV1 + 128) >> 8 W = Z - 256MVB = (W x MV1 + 128) >> 8 其中，TDB表#現行B幀與List 〇基準幀之間的時間距離， TDD，1表示list 1基準幀之攔位1與1!討〇基準攔位之間的 時間距離，MV1表示直接模式用丨丨时i基準悄之攔位i内共 位置區塊之運動向量。 以計算直接模式運動 [4 ]對圖像間時間距離值指定符號The motion vectors MVF and MVB of 0 and list 1 are as follows (Figure 133): MVF = TDB x MV1 / TDD, 1 MVB = (TDB -TDD, 1) x MV1 / TDD, 1 The above equation can be expressed as follows: Z-TDB x 256 / TDD, 1MVF = (Z x MVl + 128) > > 8 W = Z-256MVB = (W x MV1 + 128) > > 8 where TDB represents the current B frame and the Li st 0 reference frame The time distance between the two, TDD, 1 represents the time distance between the block 1 of the reference frame of list 1 and the reference block of 1 hour, and MV1 represents the total number of blocks in the list i of the reference frame of the direct mode. Motion vector. For example, since the common position block of the reference frame field i η of list 1 has a motion vector pointing to the stop of the subsequent frame in time, the 1 i st 〇 reference image is 1 ist 1 after the reference image in time. Then, calculate the direct mode Ust 〇 and list 1 motion vectors MVF and MVB as follows (Figure 13b): MVF = -TDB X MV1 / TDD, 1 MVB =-(TDB + TDD, 1) x MV1 / TDD, 1 It can be expressed as follows: Z--TDB x 25 6 / TDD, 1MVF = (Z χ MV1 + 128) > > 8 W = Z-256MVB = (W x MV1 + 128) > > 8 where TDB table #The time distance between the current B frame and List 〇 the reference frame, TDD, 1 represents the time distance between the reference frame 1 and 1 of the list 1 reference frame! 〇 〇The time frame between the reference frame and MV1 indicates the direct mode The motion vector of the co-located block in the benchmark i. Specify the sign for the time distance value between images by calculating direct mode motion [4]

1229825 五、發明說明（35) 向量之情況 若直接模式用list 1基準圖德 之後，在此情況時，可提供二上位於B圖像之前或 時間距離值指定一符號，^=异法。藉由對一圖像間 情況！ 1圖像之巨區塊與此等咖 都為幀模式 '"準圖像之共位置巨區塊 若B圖像之巨區塊與Ust !基準圖像 幀杈式時，可計算B圖像之直接模,立置巨區塊都為 下： 、連動向量MVF與MVB如1229825 V. Description of the invention (35) The case of vectors If the direct mode uses list 1 as the reference figure, in this case, you can provide a symbol that is located before the B image or the time distance value, ^ = alien method. By looking at the situation between images! 1 The huge block of the image and these other frames are in the frame mode '" The co-located giant block of the quasi-image is used if the huge block of the B image and the Ust! Reference image are framed, and the B image can be calculated. Like the direct mode, the vertical blocks are placed as follows:

MVF = TDB X MV / TDDMVF = TDB X MV / TDD

M^VB = (TDB - TDD) x MV / TDD Z = ;DB x 256 / T_F = (z χ I; = (W …128 … 八中丁DB表不一現行B幀與一Hst 離，若此距離係從B巾貞測量時，則土準^之間的時間距 此距離係從1 ist 〇基準賴測量則對日疋—正（+ )號，若 距離，若此距離係從lis"基準二0量〜準：之間的時間 ，⑴號，若此距離係從list 〇基準、'對其指定- 定—負㈠號；而訂表示一直接模式==，則對其指 共位置區塊之運動向量。輪式用llS"基準圖像内 ，況2 .·Β圖像之巨區塊與1丨以} 都為攔位模式 土+ ^像之共位置巨區塊 第39頁 1229825 五、發明說明（36) — 若B圖像之巨區塊與丨丨以1基準圖像之共位置巨.區塊都為 搁位模式時，可計算B圖像每一欄位i之直接模 A ”、、 _，i與MVB，i如下： @式運動向篁 MVF， i = TDB， i X MVi / TDD， i MVB, i = (TDB, i - TDD,i) x MVi / TDD, i 或 Z = TDB， i x 256/TDD, iMVF, i = (Z x MVi + i28) >> 8 W = z - 256MVB，i : (W x MVi + 128) >> 8 ί二距Τ’ 1表示一現行B攔位與一 11 S"基準攔位之間的 r + /味離，若此距離係從B欄位測量時，則對其指定一正 定-距^係從參1ist 0基準搁位測量時，則對其指 ^^^.,； ；；t if^^^ 時，則對J：指定一 E離係仗1 1St 1基準攔位測量 位測Μ?則對其指定)一號負 用n St I基準攔位署5儿，而MVi表示一直接模式 情況3:㈣像之區塊之運動向量。 共位置巨區塊為幀模式”、、*立拉式，而1 i s t 1基準圓像之 若β圖像之巨區塊為襴位 置巨區塊為幀模式時，钟工曾，而1 1 S t】基準圖像之共位 運動向量MVF，i與許β 斤β續每一攔位i之直接模式 :i=TH ; mT ·· MVB, i . (Τ])β i 或 TI)D) UV / m 第40頁 五、發明說明（37) z = TDB, i X 256/TDDMVF i = r?㈣M ^ VB = (TDB-TDD) x MV / TDD Z =; DB x 256 / T_F = (z χ I; = (W… 128…) The Yazhong Ding DB indicates that the current B frame is separated from an Hst, if this When the distance is measured from B., the time between the local standard and the distance is from 1 ist. The reference is measured from the sundial-positive (+) sign. If the distance is from the lis " reference, 20 amount ~ quasi: the time between, ⑴, if this distance is from the reference of list 〇, 'specify it-fixed-minus ；; and order means a direct mode ==, it refers to the co-location area The motion vector of the block. In the llS " reference image, the case 2. · Β image of the giant block and 1 丨 with} are both blocking mode soil + ^ image of the co-located giant block, page 39 1229825 V. Description of the Invention (36) — If the huge block of the B image is co-located with 1 reference image. When the blocks are in the resting mode, the direct value of each field i of the B image can be calculated. Module A ”, _, i and MVB, i are as follows: @ 式 运动 向 篁 MVF, i = TDB, i X MVi / TDD, i MVB, i = (TDB, i-TDD, i) x MVi / TDD, i or Z = TDB, ix 256 / TDD, iMVF, i = (Z x MVi + i28) > > 8 W = z-256MVB, i: (W x MVi + 128) > > 8 ί two distances T '1 represents r + / taste away between a current B stop and a 11 S " reference stop, if When this distance is measured from the B column, a positive definite distance is assigned to it-when measured from the reference 1ist 0 reference stand, it is referred to ^^^.,;;; T if ^^^, then For J: designate an E off-line 1 1St 1 reference stop measurement position M? Then specify it) No. 1 negative use n St I reference stop agency 5 children, and MVi indicates a direct mode case 3: image The motion vector of the block. The co-location giant block is in the frame mode ", **, and when the giant block of the β image in the 1 ist 1 reference circle image is the 襕 position giant block in the frame mode, Zhong Gongzeng, and 1 1 S t] the co-location motion vector MVF of the reference image, i and Xu β β β continue the direct mode of each stop i: i = TH; mT ·· MVB, i. (Τ) ) β i or TI) D) UV / m Page 40 V. Description of the invention (37) z = TDB, i X 256 / TDDMVF i = r? ㈣

Vtz，^^L=u；MV + i^ »^2δ) >>δ ^ ^ ^ ^ ^ ;st, ^ ^ ^ 號，若此距離係從丄i t η λ 、 i、扎疋一正（+) 負（-)轳基準欄位測量時，則對复指定 負（）旒，TDD表示list丨基準幀血 J日疋一 間距離，若此距離係&list準:準=之間的時 一正⑴號，若此距_心其指定 指定一負㈠於·而MV p + U &早巾貞測里時，則對其 € r* a虎， 表不一直接模式用list 1基準幀内 共位置區塊之運動向量。 丞半幀内 情=:B圖像之巨區塊為鳩模式，而 位置巨區塊為攔位模式 *旱圖像之共 若B0像之巨區塊為幀模式，而"Μ 1 巨區塊為攔位槿式，日〗· + ，# # + 口像之共位置 之德日士】 1基準圖像在時間上為Β圖像 β @ # 1 s t 1基準幀之攔位〇比其攔位1在時間上較靠近 模圖/運動；棚位0之共位置區塊的運動資訊計算直接 ’使用，Hst 1基準傾之攔位。内共 ，置&塊的運動資訊計算直接模式運動向量時，可從以 等式獲得6頓之直接模式運動向量MVF與MVB : MVF =TDB X MVO / TDD, Ο MVB = (TDB -TDD,0) χ MVO / TDD, 0 或 Z TDB χ 256/TDD,0MVF= (Z x MVO + 128) >> 8 w = z - 256MVB = (W x MVO + 128) >> 8 1229825Vtz, ^^ L = u; MV + i ^ »^ 2δ) > > δ ^ ^ ^ ^ ^; st, ^ ^ ^, if this distance is from 丄 it η λ, i (+) Negative (-) 轳 When the reference field is measured, specify negative () 复 for the complex, TDD means list 丨 the reference frame blood J day J a distance, if this distance is & list quasi: quasi = between A positive sign of the time, if this distance is assigned to a negative sign, and MV p + U & Zao Jinzheng test, then use it to list a list of direct modes. 1The motion vector of the co-located block in the reference frame.丞 Half-frame inside story =: The giant block of B image is dove mode, and the location giant block is stop mode. * The dry image is the same as the giant block of B0 image is frame mode, and " Μ 1 giant area The block is a hibiscus-style, Japanese. +, # # + The common position of the mouth image of Deritsu] 1 The reference image is temporally a B image β @ # 1 st 1 The stop of the reference frame is better than Block 1 is closer to the model / movement in time; the motion information calculation of the co-located block of booth 0 is used directly, and the Hst 1 benchmark tilts the block. When calculating the direct mode motion vector with the motion information of the internal & block, 6 direct mode motion vectors MVF and MVB can be obtained from the equation: MVF = TDB X MVO / TDD, MVB = (TDB -TDD, 0) χ MVO / TDD, 0 or Z TDB χ 256 / TDD, 0MVF = (Z x MVO + 128) > > 8 w = z-256MVB = (W x MVO + 128) > > 8 1229825

:中，TDB表示-現行"貞與list 〇基準幢之 離，若此距離係從β幀測量時，則斜发社— Τ間距 此距離係從1 i s t 0基準幀測量時，則對其指—一 &，右 號；TDD，0表示list 1基準幢之攔位Q盘二』，j (一） 日、、’，對其指疋一正（+ )號，若此距離係從丨i s t 〇基準 位測量時，則對其指定一負（一）號，·而Μν〇表示一直土 用1 i s t 1基準幀之欄位〇内共位置區塊之運動向量。、 若list 1基準圖像在時間上為B圖像之前，貝彳list i基準 幀之攔位1比其攔位〇在時間上較靠近6圖像，所以使二欄 位1之共位置區塊的運動資訊計算直接模式運動向量。因 此’使用1 i s t 1基準幀之攔位1内共位置區塊的運動資訊 計算直接模式運動向量時，可從以下等式獲得B幀之直接 模式運動向量MVF與MVB : MVF =TDB X MV1 / TDD,1 MVB = (TDB -TDD,1) χ MV1 / TDD,1 或 Z = TDB x 256/TDD,1MVF= (Z x MV1 + 128) >> 8 W = Z - 256MVB = (W x MV1 + 128) >> 8 其中’ TDB表示一現行b幀與i ist 〇基準幀之間的時間距 離，若此距離係從B幀測量時，則對其指定一正（+)號，若 此距離係從1 i s t 〇基準幀測量時，則對其指定一負（一） 號；TDD，1表示1 ist 1基準幀之欄位}與丨ist 〇基準攔位之 間的時間距離，若此距離係從1丨s t 1基準幀之欄位1測量 1229825 五、發明說------ 時’貝丨J對直扣a / 、 位測量日^/， 疋一正（+ )號’若此距離係從1 ist 0基準欄 用$ 則對其指定一負（-）號；而MV1表示一直接模式 從以基準巾貞之搁位1内共位置區塊之運動向量。 代動全$ ^顯然可知’本發明提供一種方法，用以計算次 向量了 f ^技術中定義的B(雙預測）圖像的直接模式運動 術，可择加2議_圖像直接模式運動向量之取得技 藉此改傻直接板式作為巨區塊預測模式的可能性， 、 人運Β圖像編碼效率。 此丨土 以上雖然揭示本發明之者 熟悉此類技術之人士應可理作為舉例說明之用，但 :不脫離本發明所附申請專利範‘：：作多種修改與增減 神。 圍中揭示的發明範圍與：: Medium, TDB means-the distance between the current "quotation" and the reference frame. If this distance is measured from the β frame, then the oblique hair agency — T distance This distance is measured from the 1 ist 0 reference frame. Means — one &, the right number; TDD, 0 means the stop Q disk two of the reference building of list 1 ", j (a) day, and ', refer to 疋 一 正 (+), if this distance is from丨 ist 〇 When measuring the reference position, a negative (one) number is assigned to it, and Μν〇 represents the motion vector of the co-located block within the field 0 of the 1 ist 1 reference frame. 1. If the reference image of list 1 is before the B image in time, the stop 1 of the reference frame of list i is closer than the stop 0 in time. Therefore, the common position area of the two columns 1 is set. The block's motion information calculates the direct mode motion vector. Therefore, when the direct mode motion vector is calculated using the motion information of the co-located block in block 1 of 1 ist 1 reference frame, the direct mode motion vectors MVF and MVB of frame B can be obtained from the following equation: MVF = TDB X MV1 / TDD, 1 MVB = (TDB -TDD, 1) χ MV1 / TDD, 1 or Z = TDB x 256 / TDD, 1MVF = (Z x MV1 + 128) > > 8 W = Z-256MVB = (W x MV1 + 128) > > 8 where 'TDB represents the time distance between the current b frame and the i ist 〇 reference frame. If this distance is measured from the B frame, assign a positive (+) sign to it, If this distance is measured from the 1 ist 〇 reference frame, a negative (one) number is assigned to it; TDD, 1 represents the time distance between the 1 ist 1 reference frame field} and the ist 〇 reference stop, If this distance is measured from column 1 of reference frame 1 st 1 st 1229825 V. Invented ------ When the time is measured by a pair of straight buckles a /, measurement date ^ /, 疋 一 正 （+） If this distance is from the 1 ist 0 reference column, a negative (-) sign is assigned to it; and MV1 indicates a direct mode from the motion vector of the co-located blocks in the reference frame 1 Obviously, it can be known that the present invention provides a method for calculating the secondary vector of the direct mode kinematics of the B (bi-prediction) image defined in the f ^ technique, with an optional addition of 2 _ image direct mode motion The vector acquisition technology can change the possibility of using the direct plate type as a giant block prediction mode, and the efficiency of human movement B image coding. Here, although those who have disclosed the present invention should be familiar with those who are familiar with such technology, it should be considered as an example, but it does not depart from the scope of the patent application attached to the present invention ‘: make various modifications and additions and deletions. The scope of the invention disclosed in the

12298251229825

圖式簡單說明 [圖式簡單說明] 圖la至lc顯示一般IBBBP形式之直接模式用list 像； 圖 圖2a至2c顯示一般IBBB形式之直接模式用list 像； 土 _ 圖3a至3h顯示直接模式用1 ist 1基準圖像在時間上為 圖像之後的各種情況（LO MV表示iist 〇運動向量，Li 表不list 1運動向量）； 圖4a至4h顯示直接模式用list 1基準圖像在時間上為_b 圖像之前的各種情況（LO MV表示nst 0運動向量，L1 Mv 表示list 1運動向量）； 圖5顯示考慮一般空間冗餘度而使用鄰區塊A、B、c運 量作區塊E之運動向量預測； ° 接模式用1 ist 1基 且丨ist 1基準圖像 圖6a至6c顯示一B圖像之巨區塊與一直 準圖像内共位置巨區塊都為一巾貞模式 在時間上為B圖像之後的各種情況； 圖7a至7d顯示一B圖像之巨區塊與一直接模式用i 準圖像内共位置巨區塊都為一攔位模式，且1 像在時間上為B圖像之後的各種情況； i s t 1 基 1基準圖 圖8a至8c顯示一 B圖像之巨區塊為一攔位模 模式用Hst i基準圖像内共位置巨區塊為一' llSt 1基準圖像在時間上為B圖像之後的各種:二式 ，至9。顯示-B圖像之巨區塊為一巾貞模式種γ首 式用llSt 1基準圖像内共位置巨區塊為-攔位模式，/Schematic description [Schematic description] Figures la to lc show list images for direct mode in general IBBBP format; Figures 2a to 2c show list images for direct mode in general IBBB format; Soil_ Figures 3a to 3h show direct mode The 1 ist 1 reference image is the time after the image (LO MV stands for iist 〇 motion vector, Li represents the list 1 motion vector); Figures 4a to 4h show the direct mode using the list 1 reference image at time The above are the various situations before the _b image (LO MV represents the nst 0 motion vector, and L1 Mv represents the list 1 motion vector); Figure 5 shows the use of neighboring blocks A, B, and c for the general spatial redundancy. Prediction of the motion vector of block E; ° 1 ist 1 basis and ist 1 reference image are used in the connection mode. Figures 6a to 6c show that the giant block of a B image and the co-located giant block in the quasi-image are all one. The frame pattern in time is various situations after the B image; Figures 7a to 7d show that a giant block of a B image and a co-located giant block in a quasi-image for a direct mode are both a blocking mode. And 1 image is the various situations after B image in time; ist 1 base 1 reference picture Figures 8a to 8c show the huge blocks of a B image as a blocking mode. The Hst i reference image uses a co-located giant block as a 'llSt 1 reference image in time after the B image: Second form, to 9. The giant block showing the -B image is the first pattern of the γ pattern. The macroblock with a common position in the reference image of llSt 1 is-block mode,

第44頁 1229825 圖式簡單說明 list 1基準圖像在時 圖l〇a與l〇b顯示—B圖 為B圖像之後的各種情況； 基準圖像内共位置巨區之巨區塊與一直接模式用1 i st 1 像在時間上為B圖像之^ A都為一幀模式，且1 i s t 1基準圖 圖1 1 a至1 1 d顯示一 Β圖像 ’ 基準圖像内共位置巨區靖 品束與一直接模式用list 1 圖像在時間上為b圖像之前的各種y換式且hStl基準 圖12a與12b顯示-B圖像之巨區塊為一攔位 般直接模式用list 1基準圖像内並 、式，而一 一 式，且Hst !基準圖像在時間上別圖像之;;鬼為-幢模 況；以及 各種情 圖13a與13b顯示一B圖像之巨區塊為一幀模 直接模式用list 1基準圖像内共位、'/ ，而一一般 式，且1 i s t 1基準圖像在時間上為β圖像°二、、一攔位模 況'。 為8圖像之則的各種情1229825 on page 44 A brief description of the list 1 reference image is shown in Figures 10a and 10b—B is the situation after the B image; the huge block of the co-located giant area and the Direct mode uses 1 i st 1 image in time. B image is A frame mode, and 1 ist 1 reference chart. Figures 1 a to 1 1 d show a B image. Co-location within the reference image. The giant area Jingpin beam and a direct mode use the list 1 image in time for various y transformations before the b image and hStl benchmarks Figures 12a and 12b show that the huge block of the B image is a block-like direct mode Use the union and expressions in the reference image of list 1 and the one-by-one expression, and the Hst! Reference image is different from the image in time; the ghost is a building model; and various plots 13a and 13b display a B image The giant block is a frame mode direct mode with list 1 reference image co-location, '/, and a general formula, and the 1 ist 1 reference image is β image in time. condition'. Various emotions for 8 images

第45頁Page 45

Claims (1)

1229825 /、、申清專利範圍 1 · 一種在動晝編碼系統中計算B (雙預測）圖像直接模式運 動向量以取得B圖像直接模式運動向量之方法，包括以下 步驟·從一直接模式用1 i s t 1基準圖像内共位置區塊之運 動向量中，選擇其中任一個作為用以導出該直接模式運動 向量之運動向量，不論該共位置區塊之運動向量的模式如 7(llst 0模式及/或list 1模式）；將選取之運動向量所 指向的基準圖像決定為該直接模式用丨i s t 〇基準圖像；以 及’計算該B圖像之直接模式運動向量。 2· 一胃種在動畫編碼系統中計算B(雙預測）圖像直接模式運 動向ϊ以取得B圖像直接模式運動向量之方法，包括以下 步驟·若一直接模式用1 i s t 1基準圖，内的共位置區塊只 有=個list 1運動向量，則使用該Hs]t 1運動向量作為用 以f出该直接模式運動向量之運動向量；將該list 1運動 向ΐ所指向的基準圖像決定為該直接模式用丨i s t 〇基準圖 像’以及’計异該β圖像之直接模式運動向量。 3· 一種在動晝編碼系統中計算β(雙預測）圖像直接模式運 動向量=取得Β圖像直接模式運動向量之方法，包括以下 步驟·右一直接模式用Π s t 1基準圖像内的共位置區塊只 有:個11 st 1運動向量，則視該共位置區塊有一零運動； 間上恰位於該B像之前的解碼圖像，決定為該直接模 =1曰ist 0基準圖像；以及，導出該β圖像之直接模式 動向置。 4動向一量種以在:碼系統中計算B(雙預測)圖像直接模式運 付圖像直接模式運動向量之方法，包括以下1229825 / 、 Shenqing Patent Scope 1 · A method of calculating the B (bi-prediction) image direct mode motion vector in the moving day coding system to obtain the B image direct mode motion vector, including the following steps: Among the motion vectors of the co-location blocks in the 1 ist 1 reference image, select any one of them as the motion vector for deriving the direct mode motion vector, regardless of the mode of the motion vector of the co-location block such as 7 (llst 0 mode And / or list 1 mode); determining the reference image pointed by the selected motion vector as the reference image for the direct mode; and 'calculating the direct mode motion vector of the B image. 2. A method for calculating the direct motion of the B (bi-prediction) image in the animation coding system to obtain the direct motion vector of the B image, including the following steps: If a direct mode uses a 1 ist 1 reference map, The co-location blocks within the list are only one list 1 motion vector, then the Hs] t 1 motion vector is used as the motion vector used to f out the direct mode motion vector; the list 1 is moved to the reference image pointed by ΐ It is decided to use the direct mode reference image for the direct mode and the direct mode motion vector that differentiates the β image. 3. A method for calculating β (bi-prediction) image direct mode motion vector = obtaining a direct image motion vector of a B image in a moving daytime coding system, including the following steps: The co-location block only has: 11 st 1 motion vectors, the co-location block is considered to have zero motion; the decoded image that is located immediately before the B image is determined to be the direct mode = ist 0 reference map And the direct mode of the β image is derived. (4) A method to calculate the motion vector of the direct mode of the B (bi-prediction) image in the code system, including the following: 第46頁 1229825 六、申請專利範圍 步驟：若一直接模式用1 i st 1基準圖像内的共位置區塊只 有一個list 1運動向量，則使用該共位置區塊之list 1運 動向量作為用以導出該直接模式運動向量的運動向量；將 時間上恰位在該B像之前的解碼圖像，決定為該直接模式 用list 〇基準圖像；以及，導出該β圖像之直接模式運動 向量。 5· 一種在動畫編碼系統中計算Β(雙預測）圖像直接模式運 動向量以取得Β圖像直接模式運動向量之方法，包括以下 步驟··若一直接模式用u st 1基準圖像内的共位置區塊有 二個運動向量，則從該二運動向量中選擇其中任一個 (list 0模式向量及/或list 1模式向量），並從選取的運 動向量導出該β圖像之直接模式運動向量。 6 ·如申請專利範圍第5項之方法，其中該步驟包括以下步 驟·無條件選取該1 i s t 〇運動向量作為導出該直接模式運 動向置用的運動向量，不論其時間距離如何；將該丨i st 〇 運動向置所指向的基準圖像決定為該直接模式用list 〇基 準圖像，以及’導出該β圖像之直接模式運動向量。 Ί·如申請專利範圍第5項之方法，其中該步驟包括以下步 驟·選取该list 〇與list 1運動向量之一，作為用以導出 忒直接扠式運動向量的運動向量，所選取的運動向量應指 向時間上較靠近該直接模式用丨i s t 1基準圖像的圖像；若 該二f運動向量係指向同一基準圖像，則選取該丨i s t 〇運 動向篁作為用以導出該直接模式運動向量的運動向量；將 選取之運動向量所指向的基準圖像決定為該直接模式用 1229825 申凊專利範圍 __ ^ μ 〇基準圖像；以及，導出該β圖像之直接模式運動向 g 一 番 在動晝編碼系統中計算8(雙預測）圖像直 解像直接模式運動向量…，包巧 像；制旦#被新解碼圖像設定為直接模式用1 i st 1基準圖 運動向IΤ ΐΐ模式用list 1基準圖像内的共位置區塊之 向量MVB · 出一1 iSt 0運動向量MVF與一1 ist 1運動 g 一種计异該B圖像之直接模式運動向量。 動向量取動β晝編碼系統中計算Μ雙預測）圖像直接模式運 步驟··若一 圖像直接模式運動向量之方法，包括以下 像之前時，制2模式用1 1 St 1基準圖像在時間上為該3圖 塊之運動向該直接模式用liSt 1基準圖像内共位置區 動向量MVB ，以導出一1 iSt 〇運動向量MVF與1 ist 1運 !〇·如申像之直接模式運動向量。 步驟··若該B圖#祀帛9項之方法’其中該步驟包括以下 巨區塊都為巾= 時間上為哕丨彳:w直接板式用1 i s t 0基準圖像在 接模式Λ準圖像之前時’則計算該B圖像之直 MM = 置MVF 與MVB 如下： MVR DB x MV / TDD 或 ~ (TDB ~ TDD) x MV / TDD ;:3 ;5:56 ::DMVF = (Z “v + m)〉〉8 6 MVB = (W x MV + 128) » 8 1229825 六、申請專利範圍 其中，TDB表示一現行B幀與一list 0基準幀之間的時間距 離；TDD表示一 1 i s t 1基準幀與1 i s t 0基準幀之間的時間 距離；而Μ V表示該直接模式用1 i s t 1基準圖像内共位置區 塊之運動向量。 11. 如申請專利範圍第9項之方法，其中該步驟包括以下 步驟：若該B圖像之巨區塊與該1 i s t 1基準圖像之共位置 巨區塊都為幀模式，且該直接模式用1 i s t 0基準圖像在時 間上為該1 i s t 1基準圖像之後時，則計算該B圖像之直接 模式運動向量MVF與MVB如下： MVF 二- TDB X MV / TDD MVB = - (TDB + TDD) x MV / TDD 或 Z = - TDB x 256 / TDDMVF = (Z x MV + 128) >> 8 W = Z - 256 MVB = (W x MV + 128) >> 8 其中，TDB表示一現行B幀與一list 0基準幀之間的時間距 離；TDD表示一list 1基準幀與該1 ist 0基準幀之間的時 間距離；而MV表示該直接模式用1 i st 1基準圖像内共位置 區塊之運動向量。 12. 如申請專利範圍第9項之方法，其中該步驟包括以下 步驟：若該B圖像之巨區塊該與1 i s t 1基準圖像之共位置 巨區塊都為一欄位模式，且該直接模式用1 i s t 0基準圖像 在時間上為該1 i st 1基準圖像之前時，則計算該B幀每一 攔位i之直接模式運動向量Μ V F，i與Μ V B，i如下： MVF, i = TDB, i x MVi / TDD, iPage 46 1229825 VI. Patent application steps: If a direct mode uses only a list 1 motion vector for the co-location block in the 1 i st 1 reference image, the list 1 motion vector of the co-location block is used To derive the motion vector of the direct mode motion vector; determine the decoded image immediately before the B image in time as the reference image for the direct mode list 〇; and, derive the direct mode motion vector of the β image . 5. A method of calculating a B (bi-prediction) image direct mode motion vector in an animation coding system to obtain a B image direct mode motion vector, including the following steps: If a direct mode uses the u st 1 reference image The co-location block has two motion vectors, then select any one of the two motion vectors (list 0 mode vector and / or list 1 mode vector), and derive the direct mode motion of the β image from the selected motion vector. vector. 6. The method according to item 5 of the patent application scope, wherein the step includes the following steps: Unconditionally selecting the 1 ist 〇 motion vector as the motion vector for deriving the direct mode motion direction, regardless of its time distance; The reference image pointed by st 〇 motion direction is determined as the reference image for the direct mode list 〇, and the direct mode motion vector of the β image is derived.如 The method according to item 5 of the patent application, wherein this step includes the following steps: selecting one of the list 〇 and list 1 motion vectors as the motion vector used to derive the 叉 direct fork motion vector, the selected motion vector It should point to an image that is closer to the direct mode using the ist 1 reference image in time; if the two f motion vectors point to the same reference image, select the ist 〇 motion direction 篁 as the direct mode motion The motion vector of the vector; the reference image pointed by the selected motion vector is determined as the reference image for the direct mode 1229825 patent application range __ ^ μ 〇; and, the direct mode motion of the β image is derived to g a Fan calculates 8 (bi-prediction) image direct resolution direct mode motion vector in the moving day coding system ..., encapsulating the image; Zhidan # is set by the newly decoded image to direct mode and uses 1 i st 1 reference map to move to Τ The pattern uses the vector MVB of the co-located blocks in the reference image of list 1 to produce a 1 iSt 0 motion vector MVF and a 1 ist 1 motion g. A direct mode motion vector that distinguishes the B image. Calculating M Bi-prediction in the motion vector fetching β-day coding system) Image direct mode operation steps ... If a method of image direct mode motion vector includes the following, as before, the system uses 1 1 St 1 reference image in 2 mode In time, the motion of the 3 blocks is directed to the direct mode using the co-location zone motion vector MVB in the liSt 1 reference image, to derive a 1 iSt 〇 motion vector MVF and 1 ist 1 operation! Pattern motion vector. Steps · If the B picture ## 9 method of sacrifice to the 9 items' method, where the step includes the following giant blocks are towels = time is 哕 丨 哕: w direct plate type with 1 ist 0 reference image in the connection mode As before, then calculate the straightness of the B image MM = set MVF and MVB as follows: MVR DB x MV / TDD or ~ (TDB ~ TDD) x MV / TDD;: 3; 5: 56 :: DMVF = (Z “V + m) 〉〉 8 6 MVB = (W x MV + 128)» 8 1229825 6. The scope of patent application Among them, TDB represents the time distance between an existing B frame and a list 0 reference frame; TDD represents a 1 The time distance between the ist 1 reference frame and the 1 ist 0 reference frame; and MV represents the motion vector of the co-located block in the 1 ist 1 reference image for this direct mode. 11. The method of item 9 in the scope of patent application , Where the step includes the following steps: if both the giant block of the B image and the co-located giant block of the 1 ist 1 reference image are in frame mode, and the direct mode uses the 1 ist 0 reference image in time After the 1 ist 1 reference image, the direct mode motion vectors MVF and MVB of the B image are calculated as follows: MVF 2-TDB X MV / TDD M VB =-(TDB + TDD) x MV / TDD or Z =-TDB x 256 / TDDMVF = (Z x MV + 128) > > 8 W = Z-256 MVB = (W x MV + 128) > > 8 where TDB represents the time distance between an existing B frame and a list 0 reference frame; TDD represents the time distance between a list 1 reference frame and the 1 ist 0 reference frame; and MV represents the direct mode use The motion vector of the co-located block in the 1 i st 1 reference image. 12. For the method of the 9th scope of the patent application, the step includes the following steps: If the giant block of the B image is related to the 1 ist 1 reference When all the co-located giant blocks of the image are in a one-column mode, and when the direct mode uses the 1 ist 0 reference image in time before the 1 i st 1 reference image, each block of the B frame is calculated. The direct mode motion vectors MVVF, i and MVB, i are as follows: MVF, i = TDB, ix MVi / TDD, i 第49頁 1229825 六、申請專利範圍 MVB, i =： (TDB, i ^ TDD,i) x MVi / TDD, i 或 ^ TDB, i x 256 / TDD, IMVF, i = (Z x MVi + 128) >> 8 W = Z - 256 MVB, i = (W x MVi + 128) >> 8 其中，TDB’i表系，現行B襴位與一list 0基準攔位之間的 時間距離；TDD，i表示一Hst 1基準欄位與該丨ist 〇基準 欄位之間的時間距離；而MV i表示該直接模式用丨丨st i基 準搁位内共位置區塊之運動向量。 13·如申請專利範圍第9項之方法，其中該步驟包括以下 步驟··若該B圖像之巨區塊與該1 i s t 1基準圖像之共位置 巨區塊都為一攔位模式，且該直接模式用1 i s t 〇基準圖像 在時間上為1 i s t 1基準圖像之後時，則計算一 B幀每一攔 位i之直接模式運動向量MVF，i與MVB, i如下： MVF，丨=—丁db， i X MVi / TDD, i MVB, i = . (TDB, i + TDD,i) x MVi / TDD,I 或 Z =〜TDB，i x 256 / TDD，iMVF，i = (Z x MVi + i28) >〉 8 W " Z ~ 256 MVB, i = (W x MVi + 128) >> 8 其中，TDB，i表示一現行B欄位與一list 0基準攔位之間 的時間距離；T D D，i表示一 1 i s t 1基準攔位與該i i s t 〇基 準攔位之間的時間距離；而MVi表示該直接模式用i ist 1 基準攔位内共位置區塊之運動向量。 12298251229825 on page 49 6. Scope of patent application MVB, i =: (TDB, i ^ TDD, i) x MVi / TDD, i or ^ TDB, ix 256 / TDD, IMVF, i = (Z x MVi + 128) > > 8 W = Z-256 MVB, i = (W x MVi + 128) > > 8 where TDB'i table system, the time distance between the current B position and a list 0 reference stop; TDD, i represents the time distance between an Hst 1 reference field and the ist ist 0 reference field; and MV i indicates the motion vector of the co-located block in the st i reference slot for the direct mode. 13. The method according to item 9 of the scope of patent application, wherein this step includes the following steps: If the giant block of the B image and the co-located giant block of the 1 ist 1 reference image are in a blocking mode, And when the direct mode uses a 1 ist 〇 reference image in time after the 1 ist 1 reference image, the direct mode motion vectors MVF, i and MVB, i of each stop i of a B frame are calculated as follows: MVF,丨 = —ding db, i X MVi / TDD, i MVB, i =. (TDB, i + TDD, i) x MVi / TDD, I or Z = ~ TDB, ix 256 / TDD, iMVF, i = (Z x MVi + i28) >> 8 W " Z ~ 256 MVB, i = (W x MVi + 128) > > 8 where TDB, i represents a current B field and a list 0 reference stop Time distance between the two; TDD, i represents the time distance between a 1 ist 1 reference stop and the iist 〇 reference stop; and MVi indicates that the direct mode uses the motion vectors of the co-located blocks within the i ist 1 reference stop . 1229825 步驟：1 專圖利：圍第9項之方法，其中該步驟包括以下 準圖像之二位θ署石之。巨區塊為一欄位模式，而該11 s t 1基 0基準圖像在時間上區:龙為1貞模 <，且該直接模式用1is _Β ^ 、/上為该1 1 s t 1基準圖像之前時，則計算 MVP ·、母T一攔位i之直接模式運動向量MVF，i與MVB，i如下： 5 1 -TDB, i χ MV / Ύτ)ΐ\ M^B，1 = (TDBi 1 ~TDD) X MV / TDDSteps: 1 Special profit method: The method around item 9, where this step includes the following two digits of the quasi-image θ department stone. The giant block is a one-column mode, and the 11 st 1 base 0 reference image is in the time upper zone: the dragon is 1 chastity < and the direct mode uses 1is _Β ^, / on the 1 1 st 1 reference Before the image, calculate the direct mode motion vectors MVF, i and MVB of MVP, mother T-block i, as follows: 5 1 -TDB, i χ MV / Ύτ) ΐ \ M ^ B, 1 = ( TDBi 1 ~ TDD) X MV / TDD w ^ TDB, 1 x 256 7 TDDMVF, i = (Z x MV + 128) >> 8 - 256MVB, i = (W χ MV + 128) >> 8 f中’TDB，i表示一現行b欄位與一 Hst 〇基準攔位之間 、日守間距離；TDD表示一 nst }基準幀與一 iist 〇基準幀 之間的時間距離；而Μ ν表示該直接模式用丨i s t 1基準幀内 共位置區塊之運動向量。w ^ TDB, 1 x 256 7 TDDMVF, i = (Z x MV + 128) > > 8-256MVB, i = (W χ MV + 128) > > 8 f 'TDB, i represents a current The distance between the b field and a Hst 〇 reference stop and the day guard; TDD represents the time distance between an nst} reference frame and an iist 〇 reference frame; and ν ν represents the direct mode using the ist 1 reference frame The motion vector of the inner co-location block. 15.如申請專利範圍第9項之方法，其中該步驟包括以下 步驟··若該B圖像之巨區塊為一攔位模式，而該1 i st 1基 準圖像之共位置巨區塊為一幀模式，且該直接模式用1 i st 〇基準圖像在時間上為該丨i st 1基準圖像之後時，則計算 一 B幀每一欄位i該直接模式運動向量Μ V F，i與Μ V B，i如下： MVF, i = -TDB, i x MV / TDD MVB,i = - (TDB,i +TDD) x MV / TDD 或 2 = -TDB, i x 256 / TDD,iMVF,i = (Z x MVi + 128) >> 815. The method according to item 9 of the scope of patent application, wherein the step includes the following steps: If the giant block of the B image is a blocking mode, and the co-located giant block of the 1 i st 1 reference image Is a one-frame mode, and the direct mode uses the 1 i st 〇 reference image to be temporally after the 丨 i st 1 reference image, then calculate the direct mode motion vector MV VF of each field i of a B frame, i and MV VB, i are as follows: MVF, i = -TDB, ix MV / TDD MVB, i =-(TDB, i + TDD) x MV / TDD or 2 = -TDB, ix 256 / TDD, iMVF, i = (Z x MVi + 128) > > 8 1229825 ------ 六、申請專利範圍 …中丁DB，1表不一現行β攔位與一list n其淮辆 時間距離，· TDD表示一 1 i s t j A進結 .土準欄位之間的 間的時間距離，·而MV表示該直土接模Y、 1 1 St 〇基準幀之 位置區塊之運動向量。 接模式用〗ist】基準幀内共 16.如申請專利範圍第9項之方法，A 步驟：若該B圖像之£區塊為一幢模^ ^驟包括以下 圖像之共位置巨區塊為一攔位模、，工\而6亥1 ist 1基準 式用1 i st 0基準圖像在時間上為;’·且該B圖像之直接模 時，則從以下等式計算—B幢之直接1^ 1基準圖像之前 MVB ’其中使用一 list i基準鴨:式運動向量MV:與 動資訊，計算該直接模式運動向量Π 1内共位置£塊之運 MVF =TDB X MV1/TDD,1 ^ MVB = (TDB -TDD,1) x MVl/TDD,i WZ = ZTDB =:/ T:D’1MVF= (Z X MV…28) >> 8 :V = MV1 +l28) >> 8 其中，TDB表不一現行b ψ自盥1 . 距離；TDDJ表示該llst ^、料⑴〇基準賴之間的時間 欄位之間㈣間距離；而=貞；攔位1與：llst °基/ ，-隹e ^MV1表不該直接模式用list 1基 準幀之欄位1内共位置區塊之運動向旦。 17.如申請專利範圍第9項之方法，；中該步驟包括以下 像Λ巨區塊為1模式，而該Hs"基準 • ’、 °〇°丨為一攔位模式，且該B圖像之直接模 第52頁 1229825 六、申請專利範圍 式一 1 i s t 〇基準圖像在時間上為丨丨s t 1基準圖像之後時， 則從以下等式計算一B幀之直接模式運動向量MVF與MVB， 其中使用一 1 ist 1基準幀之攔位1内共位置區塊之運動資 訊’計算該直接模式運動向量： (TDB + TDD, 1) x MV1 / TDD, 1 MVF = -TDB X MV1 / TDD, 1 MVB =. 或 其中 TDB x 256 / TDD，1MVF= (z X MV1 + 128) >> 8 - 256MVB = (W x MV1 + 128) >> 8 T D B表示一現行B幀與一 1 i s t 0基準幀之間的時間距 離’ TDD，1表示一 1 i st 1基準幀之攔位1與一 1 i st 0基準攔 位之間的時間距離；而Μ V1表示該直接模式用丨i s t 1基準 十貞之攔位1内共位置區塊之運動向量。 18· 一種在動畫編碼系統中計算B(雙預測）圖像直接模式 運動向量以取得B圖像直接模式運動向量之方法，包括以 下步驟··若直接模式用1 i s t 〇基準圖像與丨丨s t 1基準圖像 在時間上都為該B圖像之後時，測量該直接模式用丨丨Μ丄 ，準圖像内共位置區塊之運動向量，以導alist 〇運動向 量MVF與list i運動向tMVB ;並計算該Β圖像之直 ° 運動向量。 丧棋式 19,如申請專利範圍第18項之方法，其中該步驟包括以 步驟：若該β圖像之巨區塊與該list i基準圖像之共 為巾貞模式’且該直接模式用lis"基準圖、 $上為§亥11 st 1基準圖像之後時，則計算該B圖像之直接1229825 ------ 6. Scope of patent application ... Zhongding DB, 1 indicates the time distance between the current β stop and a list n, and · TDD indicates that 1 istj A enters the knot. The time distance between the two, and MV represents the motion vector of the position block of the direct soil joint Y, 1 1 St 0 reference frame. For connection mode ist] 16. There are a total of 16. In the reference frame, the method of item 9 of the scope of the patent application, step A: If the £ block of the B image is a model ^ ^ Step includes the following image of the co-location macro area The block is a blocking mode, and the 6i 1 ist 1 reference formula uses the 1 i st 0 reference image in time; '· and the direct mode of the B image is calculated from the following equation— Before the direct 1 ^ 1 reference image of block B, MVB 'which uses a list i reference duck: type motion vector MV: and motion information to calculate the direct position of the direct mode motion vector Π 1 in total £ MVF = TDB X MV1 / TDD, 1 ^ MVB = (TDB -TDD, 1) x MVl / TDD, i WZ = ZTDB =: / T: D'1MVF = (ZX MV… 28) > > 8: V = MV1 + l28) > > 8 where TDB represents a current b ψ self-washing 1. distance; TDDJ represents the distance between the time fields between the llst ^ and the material reference 0; and = =; stop 1 And: llst ° base /,-隹 e ^ MV1 indicates that the direct mode should use the motion of the co-located blocks in column 1 of list 1 reference frame. 17. The method according to item 9 of the scope of patent application; in this step, the following steps include that the Λ giant block is in the 1 mode, and the Hs " benchmark • ', ° 〇 ° 丨 is a blocking mode, and the B image Direct mode on page 52 1229825 VI. Patent application range Formula 1 ist 〇 When the reference image is in time 丨 丨 st 1 reference image, then calculate the direct mode motion vector MVF of a B frame from the following equation MVB, which uses the motion information of a co-location block in Block 1 of a 1ist 1 reference frame to calculate the direct mode motion vector: (TDB + TDD, 1) x MV1 / TDD, 1 MVF = -TDB X MV1 / TDD, 1 MVB =. Or TDB x 256 / TDD, 1MVF = (z X MV1 + 128) > > 8-256MVB = (W x MV1 + 128) > > 8 TDB means an existing B frame and A time distance between 1 ist 0 reference frames' TDD, 1 represents the time distance between 1 1 st frame 1 reference frame and 1 1 st 0 frame reference; and M V1 indicates the direct mode丨 ist 1 The motion vector of the co-located block in the 1st position of the base 10. 18 · A method of calculating a B (bi-prediction) image direct mode motion vector in an animation coding system to obtain a B image direct mode motion vector, including the following steps: · If the direct mode uses 1 ist 〇 reference image and 丨 丨When the st 1 reference image is after the B image in time, measure the direct vector using the motion vector of the co-located blocks in the quasi-image to guide the motion of the alist 〇 motion vector MVF and list i To tMVB; and calculate the straight ° motion vector of the B image. Funeral style 19, such as the method of claim 18 in the patent application range, wherein the step includes the step of: if the large block of the β image and the reference image of the list i are in the chastity mode 'and the direct mode is used lis " When the reference image, $ §11 is after the st 1 reference image, the direct calculation of the B image is calculated 1229825 六、申請專利範圍 模式運動向量MVF與MVB如下 MVF MVB 或 TDB X MV / TDD (TDB - TDD) X MV / TDD Z = TDB x 256 / TDDMVF = (Z x MV + 128) >> 8 W = Z - 256MVB = (W x MV + 128) >> 8 其中，TDB表示一現行B幀與一 1 i s t 0基準幀之間的時間 距離；TDD表示一list 1基準幀與該list 0基準幀之間的 時間距離；而MV表示該直接模式用1 i s t 1基準圖像内共位 置區塊之運動向量。 20.如申請專利範圍第1 8項之方法，其中該步驟包括以下 步驟：若該B圖像之巨區塊與該1 i s t 1基準圖像之共位置 巨區塊都為一幀模式，且該直接模式用1 i s t 0基準圖像在 時間上為該1 i st 1基準圖像之前時，則計算該B圖像之直 接模式運動向量MVF與MVB如下： MVF = - TDB X MV / TDD MVB = - (TDB + TDD) X MV / TDD 或 Z = - TDB x 256 / TDDMVF = (Z x MV + 128) >> 8 W = Z - 256MVB = (W x MV + 128) >> 8 其中，TDB表示一現行B幀與一 1 i s t 0基準幀之間的時間 距離；TDD表示一list 1基準幀與1 i st 0基準幀之間的時 間距離；而MV表示一直接模式用1 i st 1基準圖像内共位置 區塊之運動向量。1229825 VI. Patent application range mode motion vectors MVF and MVB are as follows: MVF MVB or TDB X MV / TDD (TDB-TDD) X MV / TDD Z = TDB x 256 / TDDMVF = (Z x MV + 128) W = Z-256MVB = (W x MV + 128) > > 8 where TDB represents the time distance between an existing B frame and a 1 ist 0 reference frame; TDD represents a list 1 reference frame and the list 0 The time distance between the reference frames; and MV represents the motion vector of the co-located block in the 1 ist 1 reference image for this direct mode. 20. The method according to item 18 of the scope of patent application, wherein the step includes the following steps: if both the giant block of the B image and the co-located giant block of the 1 ist 1 reference image are in a frame mode, and When the direct mode uses a 1 ist 0 reference image in time before the 1 i st 1 reference image, the direct mode motion vectors MVF and MVB of the B image are calculated as follows: MVF =-TDB X MV / TDD MVB =-(TDB + TDD) X MV / TDD or Z =-TDB x 256 / TDDMVF = (Z x MV + 128) > > 8 W = Z-256MVB = (W x MV + 128) > > 8 Among them, TDB indicates the time distance between an existing B frame and a 1 ist 0 reference frame; TDD indicates the time distance between a list 1 reference frame and 1 i st 0 reference frame; and MV indicates a direct mode using 1 i st 1 The motion vector of the co-located block in the reference image. 第54頁 Ϊ229825 “、申請專利範圍 』·驟如!請專利範圍第18項之方法中該步驟包括以下 右该B圖像之巨區塊與該Hst i基準圖像之共 :鬼都為一攔位模式，且該直接模式用丨丨 攔位i 為該11 st 1基準圖像之後時，則計算一B幀每— Mv 之直接模式運動向量MVF，i與MVB, i如下： _，i 〜ΜΠ / ，丨 ^ UDB, 1 ^ TDD, i) x MVi / TDD, iPage 54 Ϊ229825, “Scope of Patent Application” · Such as! In the method of item 18 of the patent scope, this step includes the following: the huge block of the B image and the Hst i reference image: the ghost is one When the blocking mode is used and the direct mode is used after the 11 st 1 reference image, the direct mode motion vectors MVF, i and MVB, i of each B-frame Mv are calculated as follows: _, i ~ ΜΠ /, 丨 ^ UDB, 1 ^ TDD, i) x MVi / TDD, i •1 x 256 / TDD,IMVF,i = (Z x MVi + 128) >> W = 7 256MVB, i = (W x MVi + 128) >> 8 厂中，TDB，i表示一現行請位與一Hst 〇基準 準:間距離；TDD，i表示一list 1基準欄位與該list 装、f位之間的時間距離；而MVi表示該直接模式用iist j 卷準攔位内共位置區塊之運動向量。 ^ ·如申凊專利範圍第1 8項之方法，其中該步驟包括以下 乂，·若該B圖像之巨區塊與該1 i s t 1基準圖像之共位置 巨1塊都為一欄位模式，且該直接模式用1 i s t 0基準圖像 在日守間上為該1 i s t 1基準圖像之前時，則計算一 b幀每一 才闌位1之直接模式運動向量MVF，i與MVB，i如下： MVF， 1 = _ TDB， i X MVi / TDD， i MVB，1 = - (TDB，i + TDD，i) x MVi / TDD，I 或 Z = ~TDB, i x 256/TDD, iMVF, i = (Z x MVi + 128) >> 8• 1 x 256 / TDD, IMVF, i = (Z x MVi + 128) > > W = 7 256MVB, i = (W x MVi + 128) > > 8 In the factory, TDB, i represents a current Please refer to the distance between a Hst 〇 reference level; TDD, i represents the time distance between a list 1 reference field and the list and f positions; and MVi indicates that the direct mode uses the iist j standard stop position. The motion vector of the location block. ^ · The method of applying item 18 of the patent scope, wherein the step includes the following steps: If the giant block of the B image and the giant block of the 1 ist 1 reference image are in the same column When the direct mode uses the 1 ist 0 reference image in front of the 1 ist 1 reference image on the day, the direct mode motion vectors MVF, i, and MVB of each appendix 1 in a b frame are calculated. , I is as follows: MVF, 1 = _ TDB, i X MVi / TDD, i MVB, 1 =-(TDB, i + TDD, i) x MVi / TDD, I or Z = ~ TDB, ix 256 / TDD, iMVF , i = (Z x MVi + 128) > > 8 第55頁 1229825 申請專利範圍 w =z - 256MVB,i = cw x M 其中，丁DB，i 表示一 — t >> 8 的時間距離；現^丁j攔位與一1 ist 〇基準攔位之間 準欄位之間的時間1離1二t」基準攔位與該list 〇基 rr::共位置區：之=1 向量示㈣ 步驟：若該B專圖，象乾之圍第』8項之方法，其中該步驟包括以下 準圖像之共位置^為1位模 <，而該list !基 〇基準圖像在時間上為巾貞模式，且該直接模式用list -B t貞每一攔位i之直為接模1St 1基準圖像之後時’則計算 MVF, i =TDB ： v XMr 、式運動向量MVF, i 與MVB, i 如下·· MV / TDD / TDD =(Z X MV 128) >> 或，1 = (TDB， 1 —TDD) x MV 128) >> l ^ TDB,1 x 256 / TDDMVF, i w " z - 256MVB, i = rw Y mv 其中，tdb ，一1 (W X MV + 128) » 8 之間的時間距離.ilst i基準幀與一1ist 〇基準幀 3位置區塊之運動:量該直接模式用list 1基犧 準圖像：=像之严區塊為-攔位模式，而該基 0基準圖傻、t拉μ巨區塊為一幀模式，且該直接模式用1 i s1 —B幀每一欄位.^上為該jist 1基準圖像之前時，則計算 、 1之直接模式運動向量MVF，i與MVB，i如下： 1229825 六、申請專利範圍 MVF, i = -TDB, i X MV / TDD MVB, i = ^ (Tdb5 i +TDD) χ My f TD]) 或 Z = -TDB,i x 256/TDDMVF,i = (Z x MV + 128) >> 8 W = Z - 256MVB, i = (W x MV + 128) » 8 其中，TDB’i表示一現行^闌位與一Hst 〇基準攔位之 時間距離；TDD表示一list !基準幀與一list 〇基準幀之 間的時間距離；表示該直接模式用list丨基準巾貞内丘 位置區塊之運動向量。 、/、 2 5.如=明專利範圍第1 8項之方法，其中該步驟包括以下 步驟：若該B圖像之巨區塊為一幀模式，而該Hst i基準 圖像之/、位置巨^塊為一搁位模式，且該直接模式用1 i s t 〇基準圖像，在時間上為該丨丨以1基準圖像之後時，則從以 下等式計算一Β幀之直接模式運動向4MVF與"8，其中使 用一 1 i s t 1基準幀之攔位〇内共位置區塊之運動資訊， 鼻該直接模式運動向量： MVF =TDB χ MVO / TDD，0 MVB = (TDB -TDD, 〇) x MVO / TDD, 〇 或 Z = TDB x 256 / TDD，0MVF= (Z x MVO + 128) >> 8 W = Z - 256MVB = (w χ MVO + 128) >> 8 其中’TDB表示一現行b幀與一 list 〇基準幀之間的時間距 離；TDD，0表示該list 1基準幀之攔位〇與一list 〇基準棚 位之間的時間距離；而MV0表示該直接模式用nst }基準 1229825 六、申請專利範圍 悄之欄= 内共位置區塊之運動向量。 步驟：若二B專圖利傻耽圍第1 8項之方法，其中該步驟包括以下 丄=置圖 下笨$蚪管為4 1 1基準圖像之前時，則從以 用-l'Jt 7其、隹、之直接模式運動向量MVF與MVB ’其中使 内共位置區塊之運動資訊，計 MVF = -TDB X MVO / TDD,0 MVB = - (TDB+TDD,〇) x MV〇 / TDD 〇 或 ’ :=:TDB x 2 5 6/TDD，〇MVF= (z χ mv" i28)〉” - Z - 2 56M一VB = (w x MVO + 128)〉> 8 i · τ二，ϊ T >現订B巾貞與一list 0基準+貞之間的時間距 ,，表不該llst 1基準幀之攔位〇與一 ust 〇基準攔 巾貞之欄位0内共位置區塊之運動向量。 2J· 一種在動晝編碼系統中計算B(雙預測）圖像直接模式 運動向量以取得B圖像直接模式運動向量之方法，包括以 下步驟：對一圖像間時間距離值指定一符號；測量直 f用Ust 1基準圖像内共位置區塊之運動向量，不論直接 杈式用list 0基準圖像與該list }基準圖像之位置如何， 以V出list 0運動向量MVF與1 ist i運動向量Μνβ ;並 該B圖像之直接模式運動向量。 ° ^1229825 on page 55. The scope of patent application w = z-256MVB, i = cw x M where Ding DB, i represents a time distance of t —> t> 8; the current position of Ding j and a reference of 1 ist 〇 The time between the quasi-columns between the positions 1 to 1 2 t "The reference stop and the list 〇Base rr :: Co-location area: = 1 = Vector display ㈣ Step: If the B picture, Xiangganweiwei The method of item 8, wherein the step includes the following common position of the quasi-image ^ is a 1-bit mode < and the list! Base 0 reference image is in temporal mode, and the direct mode uses list -B Straight to each stop i is the time after the reference image of the 1St 1 die, then calculate MVF, i = TDB: v XMr, the formula motion vectors MVF, i and MVB, i are as follows ... MV / TDD / TDD = (ZX MV 128) > > or, 1 = (TDB, 1 —TDD) x MV 128) > > l ^ TDB, 1 x 256 / TDDMVF, iw " z-256MVB, i = rw Y mv Among them, tdb, the time distance between 1 (WX MV + 128) »8. The motion of the 3 position block of the ilst i reference frame and a 1ist 〇 reference frame: the direct mode uses the list 1 basis to sacrifice the image: = Strict block like-block mode The base 0 reference map is silly, t pulls μ huge blocks into a frame mode, and the direct mode uses 1 i s1 —B frame for each field. ^ Is the time before the jist 1 reference image, then calculate, The direct mode motion vectors MVF, i and MVB, 1 are as follows: 1229825 VI. Patent application scope MVF, i = -TDB, i X MV / TDD MVB, i = ^ (Tdb5 i + TDD) χ My f TD]) Or Z = -TDB, ix 256 / TDDMVF, i = (Z x MV + 128) > > 8 W = Z-256MVB, i = (W x MV + 128) »8 where TDB'i represents a current ^ The time distance between the appendix and a Hst 〇 reference stop; TDD represents the time distance between a list! Reference frame and a list 〇 reference frame; indicates the movement of the block in the direct mode using the list vector. 、 /, 2 5. If = the method of item 18 of the patent scope, this step includes the following steps: if the huge block of the B image is a frame mode, and the / position of the Hst i reference image The giant block is a shelving mode, and the direct mode uses a 1 ist 〇 reference image. In time, after the reference image, the direct mode motion direction of a B frame is calculated from the following equation. 4MVF and " 8, where the motion information of a common location block within 0 1 ist 1 reference frame is used, and the direct mode motion vector is: MVF = TDB χ MVO / TDD, 0 MVB = (TDB -TDD, 〇) x MVO / TDD, 〇 or Z = TDB x 256 / TDD, 0MVF = (Z x MVO + 128) > > 8 W = Z-256MVB = (w χ MVO + 128) > > 8 where 'TDB represents the time distance between an existing b frame and a list 〇 reference frame; TDD, 0 represents the time distance between the stop 0 of the list 1 reference frame and a list 〇 reference booth; and MV0 represents the direct The mode uses nst} benchmark 1229825. 6. The column for the scope of patent application = the motion vector of the internal common location block. Step: If the second B special method is used to cover the eighteenth method, the step includes the following steps: Set the image below the stupid $ 蚪 tube before the 4 1 1 reference image, then use -l'Jt 7 Its, 隹, and its direct mode motion vectors MVF and MVB 'where the motion information of the internal co-location block is calculated as MVF = -TDB X MVO / TDD, 0 MVB =-(TDB + TDD, 〇) x MV〇 / TDD 〇 or ': =: TDB x 2 5 6 / TDD, 〇MVF = (z χ mv " i28)〉 ”-Z-2 56M-VB = (wx MVO + 128)〉 > 8 i · τ, ϊ T > The time interval between the B frame and a list 0 reference + frame is now set, which indicates that the llst 1 frame of reference frame 0 and a ust 〇 reference frame of frame 0 in the common position block 2J · A method for calculating the direct mode motion vector of a B (bi-prediction) image in a moving day coding system to obtain the direct mode motion vector of a B image, including the following steps: Specifying a time distance value between images Symbol; measure the motion vector of the co-location block in the Ust 1 reference image for direct f, regardless of the position of the direct branch list 0 reference image and the list} reference image, V lists 0 motion vectors MVF and 1 ist i motion vectors Μνβ; and direct mode motion vectors of the B image. ° ^ 1229825 六、申請專利範圍 28. 如申請專利範圍第27項之方法，其中該步驟包括以下 步驟：若該B圖像之巨區塊與該1 i s t 1基準圖像之共位置 巨區塊都為一幀模式，則計算該B圖像之直接模式運動向 量MVF與MVB如下： MVF = TDB X MV / TDD MVB = (TDB - TDD) x MV / TDD 或 Z = TDB x 256 / TDDMVF 二（Z x MV + 128) >> 8 W 二 Z - 256MVB = (W x MV + 128) >> 8 其中，T D B表示一現行B幀與一 1 i s t 0基準幀之間的時間距 離，若此距離係從B幀測量時，則對其指定一正（+ )號，若 此距離係從1 i st 0基準幀測量時，則對其指定一負（-） 號；TDD表示一 1 i s t 1基準幀與1 i s t 0基準幀之間的時間 距離，若此距離係從1 i s t 1基準幀測量時，則對其指定一 正（+ )號，若此距離係從1 i s t 0基準幀測量時，則對其指 定一負（-）號；而MV表示該直接模式用list 1基準圖像内 共位置區塊之運動向量。 29. 如申請專利範圍第27項之方法，其中該步驟包括以下 步驟：若該B圖像之巨區塊與該1 i s t 1基準圖像之共位置 巨區塊都為一欄位模式，則計算一 B幀每一欄位i之直接模 式運動向量MVF，i與MVB, i如下： MVF, i = TDB, i x MVi / TDD, i MVB, i = (TDB, i - TDD, i) x MVi / TDD, i 或1229825 VI. Application for Patent Scope 28. For the method of applying for Patent Scope Item 27, this step includes the following steps: If the giant block of the B image and the co-located giant block of the 1 ist 1 reference image are both For one frame mode, the direct mode motion vectors MVF and MVB of the B image are calculated as follows: MVF = TDB X MV / TDD MVB = (TDB-TDD) x MV / TDD or Z = TDB x 256 / TDDMVF two (Z x MV + 128) > > 8 W 2 Z-256MVB = (W x MV + 128) > > 8 where TDB represents the time distance between an existing B frame and a 1 ist 0 reference frame, if this When the distance is measured from the B frame, a positive (+) sign is assigned to it. If the distance is measured from the 1 i st 0 reference frame, a negative (-) sign is assigned to it; TDD means a 1 ist 1 The time distance between the reference frame and the 1 ist 0 reference frame. If this distance is measured from the 1 ist 1 reference frame, assign a positive (+) sign to it. If this distance is measured from the 1 ist 0 reference frame, , A negative (-) sign is assigned to it; and MV represents the motion vector of the co-located block in the reference image of list 1 for this direct mode. 29. For the method of claim 27 in the scope of patent application, wherein the step includes the following steps: If both the giant block of the B image and the co-located giant block of the 1 ist 1 reference image are in a column mode, then Calculate the direct mode motion vectors MVF, i and MVB, i of each field i of a frame B as follows: MVF, i = TDB, ix MVi / TDD, i MVB, i = (TDB, i-TDD, i) x MVi / TDD, i or 128) >> 六、申請專利範圍 ;:x 256/TDD iMVF i = (z X … 256MVB, i = (W x MVi + 128) xN 〇 ^中，TDB，i表示一現行^闌位與一 h 攔位之間的時間距離，：此3丄基準襴位與lis"基準 位測量時，則對其指定一負（_) 攸llst 〇基準攔 用基準搁位内共位】(區 30.如申請專利範圍第27項之方法，其中該牛 步驟：若該B圖像之巨區塊為一攔位模式，γ琴匕 準圖像之共位置巨區塊為一幀模式 ：：s : 1 ^ ^ ^ t, ^ *MVF, , ^MVB, i 1：7： ^ MVF, i =TDB, i x MV / TDD MVB, i - (TDB, i -TDD) x MV / TDD 或 Z = TDB, i X 256/TDDMVF,i = (Z x MV + i28) >> 8 W = Z - 256MVB, i = (W x MV + 128) » 8 其中，TDB 乂表示一現行B棚位與一 ust 〇基準爛位之間的 時間距離，若此距離係從3欄位測量時，則對其指定一正 (+ )號，若此距離係從1 i s t 〇基準攔位測量時， 定一負（-)號JDD表示一 list }基準幢與_Ust、〇基準鴨 之間的時間距離，若此距離係從丨i st i基準幀測量時，則 1229825 六、申請專利範圍 對其指定一正（+ )唬，若此距離係從i i s t 〇基準幀測量 時，則對其指定一負（-）號；而"表示該直接模式用丨is 1基準幀内共位置區塊之運動向量。 31·如申請專利範圍第27項之方法，其中該步驟包括以 步驟：若該B圖像之巨區塊為一幀模式，而該Hst !美 圖像之共位置巨區塊為一攔位模式，且該Hst i基準^ 在時間上為該B圖像之後時，則從以下等式計算_β巾貞I 接模式運動向量MVF與MVB，其中使用一Hst 1基準巾貞= ，〇内共位置區塊之運動資訊，計算該直接模式運動向索 量： MVF =TDB X MVO / TDD, Ο MVB = (TDB -TDD’O) χ MVO / TDD Ο 或 ， TDB x 256 TDD,0MVF= (Z x MVO W 豆- Z - 256MVB = (W x MVO + 128) >> 8 ί中若ΤΓ表示一現则與-1 is t 〇基準巾貞之間的時間卵 此π μ匕距離係從B幀測量時，則對其指定一正（+ )铲.. 政匕距離係&list 〇基準鴨測量時， （+ )虓，若 號；TDD η本μ · 幻訂具才曰疋一負 之間的時門H1St 1基準幀之欄位0與—Ust 0基準攔位 糾A、/對其指定一正（+ )號，若此距離传；Τ Π + η * 攔位測量時，則 係仗list 0基準 式用1 i S t 1美里結八二 、〜，而Μ V0表示該直接模 32.如中之欄位G内共位置區塊之運動向量。 專利純圍第27項之方法’其中該步驟包括以下 1229825 六、申請專利範圍 步驟：若該B圖像之巨區塊為一幀模式，而該nst 1基準 圖像之共位置巨區塊為一欄位模式，且該丨i s t 1基準圖像 在時間上為或B圖像之前時’則從以下等式計算一 B巾貞之直 接模式運動向量MVF與MVB，其中使用一list° Γ基準幀之攔 位1内共位置區塊之運動資訊’計算該直接模式運動向 量： MVF =TDB X MV1 / TDD, 1 MVB = (TDB -TDD, 1) χ MV1 / TDD, 1 或 z = TDB X 256 / TDD,1MVF= (Z χ MVi + 128) » 8 w = Z - 256MVB = (W X MV1 + 128) >> 8 f中’TDB表示-現行B幀與一 Hs"基準幀之間的時間距 離，若此距離係從B幀測量時，則對其指定一正 此距離係從丨ist 〇基準幀測量時，則對其指定一負（:） 唬，TDD，1表示該list】基準幀之攔位〗 〇 位之間的時間距離，若此距離係從u 測量時，則對其指定一正⑴號，若；=攔= 準欄位測量時，則對其指 〇基 模式用list 1基準幀之攔^ ^虎’而MV1表不該直接 〇〇 一錄户如t早幀之欄位1内共位置區塊之運動向量。 還動/ 旦編碼系統令計算β(雙預測）圖像直接模犬 3;”取:β圖像直接模式運動向量之方 下步驟·右一直接模式用 L括以 為一，在内，模,Β本t 1基卓圖像内共位置巨區塊 、式寺’則以空間冗餘度為基礎，從該B圖像 1229825 六、申請專利範圍 之欲編碼巨區塊的鄰區塊預測並計算1 i s t 0基準圖像與 list 1基準圖像及運動向量，並計算該B圖像之直接模式 運動向量。 3 4.如申請專利範圍第3 3項之方法，其中該步驟包括以下 步驟：若該欲編碼巨區塊之鄰區塊A、B、C係表示不同的 基準圖像時，則選取索引最小之基準圖像，作為每一 1 i s t 之基準圖像。 3 5.如申請專利範圍第3 3項之方法，其中該步驟包括以下 步驟：若該欲編碼巨區塊之鄰區塊中，有二或多個係表示 索引相同之基準圖像時，則選取該基準圖像作為每一 1 i s t 之基準圖像。 36. 如申請專利範圍第33項之方法，其中該步驟包括以下 步驟：若該欲編碼巨區塊之鄰區塊A、B、C中，有任一個 為’在内’模式時，將其list 0與list 1運動向量設定為 0 ;選取一運動向量，其方向須與每一 list之基準圖像從 一鄰區塊起算之時間位置的方向相同；以及，經由中值運 算取得每一 list之運動向量；或者，若一鄰區塊有二個同 方向之運動向量時，則從該區塊只選取二個向量之一，並 經由包括所選取運動向量之中值運算，取得每一 list之運 動向量。 37. 如申請專利範圍第33項之方法，其中該步驟包括以下 步驟：若無法導出每一 1 i s t模式之有效基準圖像索引時， 則將list 0基準圖像索引與list 1基準圖像索引設定為 0，並將每一 list模式之運動向量設定為0。(128) > > Sixth, the scope of patent application ;: x 256 / TDD iMVF i = (z X… 256MVB, i = (W x MVi + 128) xN 〇 ^, TDB, i represents a current ^ appendix and The time distance between the h stops: When the 3 丄 reference position is measured with the lis " reference position, a negative (_) is assigned to it. 〇llst 〇 The reference stop is in common position in the reference position] (area 30 The method according to item 27 of the scope of patent application, wherein the cattle step: if the giant block of the B image is in a blocking mode, the co-located giant block of the gamma image is a frame mode :: s : 1 ^ ^ ^ t, ^ * MVF,, ^ MVB, i 1: 7: ^ MVF, i = TDB, ix MV / TDD MVB, i-(TDB, i -TDD) x MV / TDD or Z = TDB , i X 256 / TDDMVF, i = (Z x MV + i28) > > 8 W = Z-256MVB, i = (W x MV + 128) »8 where TDB 乂 represents a current B-shed and a ust 〇 The time distance between the reference positions. If this distance is measured from 3 columns, a positive (+) sign is assigned to it. If this distance is measured from the 1 ist 〇 reference stop, a negative (-) JDD indicates the time between a list} reference building and _Ust, 〇 reference duck Distance, if this distance is measured from the reference frame of i st i, then 1229825 VI. Assign a positive (+) sign to the patent application range. If this distance is measured from the iist 〇 reference frame, specify 1 A minus (-) sign; and " indicates that the direct mode uses the motion vector of the co-located block in the reference frame of 31. 31. For example, the method of item 27 of the scope of patent application, wherein the step includes the step: if the B The giant block of the image is a frame mode, and the co-located giant block of the Hst! Beauty image is a blocking mode, and when the Hst i reference ^ is in time after the B image, then from the following The equation calculates _β towel I connected with the pattern motion vectors MVF and MVB. Among them, the motion information of a common location block within a Hst 1 reference towel is used to calculate the amount of direct mode motion: MVF = TDB X MVO / TDD, Ο MVB = (TDB -TDD'O) χ MVO / TDD 〇 Or, TDB x 256 TDD, 0MVF = (Z x MVO W Bean-Z-256MVB = (W x MVO + 128) > > 8 If ΤΓ represents the time between a present and -1 is t 〇 reference frame, this π μ distance is measured from the B frame, then Its designation is a positive (+) shovel .. The distance of the political dagger is & list 〇 When the reference duck is measured, (+) 虓, if the number; TDD η this μ · The magic book is called the time gate between 疋 and negative. 1 The reference frame field 0 and —Ust 0 reference frame correction A, / assigned a positive (+) sign to it, if this distance is transmitted; Τ Π + η * when the measurement of the frame, the list 0 reference formula Use 1 i S t 1 Miri knot 82, ~, and M V0 represents the motion vector of the co-location block in the direct module 32. The 27th method of the patent pure circle, wherein the step includes the following 1229825 VI. Patent application step: If the giant block of the B image is a frame mode, and the common location giant block of the nst 1 reference image is One-column mode, and the ist 1 reference image is in time or before the B image, then the direct mode motion vectors MVF and MVB of a B frame are calculated from the following equations, where a list ° Γ reference frame is used The motion information of the common location block in Block 1 'calculates the direct mode motion vector: MVF = TDB X MV1 / TDD, 1 MVB = (TDB -TDD, 1) χ MV1 / TDD, 1 or z = TDB X 256 / TDD, 1MVF = (Z χ MVi + 128) »8 w = Z-256MVB = (WX MV1 + 128) > > 8 f 'TDB means-the time between the current B frame and an Hs " reference frame Distance, if the distance is measured from the B frame, a positive value is assigned to it; when measured from the reference frame, a negative (:) is assigned to it, TDD, 1 means the list] reference frame Blocking time: The time distance between 〇, if this distance is measured from u, a positive sign is assigned to it, if; = blocking = In the quasi-field measurement, it refers to the use of the list 1 reference frame in the base mode ^ ^ Tiger ', and MV1 should not directly indicate the movement of a co-located block in the field 1 in the early frame of t. vector. The motion / density encoding system calculates the β (bi-prediction) image direct model 3; "Take: the next step of the β image direct mode motion vector. The right direct mode is enclosed by L, and the module, Based on the spatial redundancy, the co-located giant blocks and style temples in the base image of t1 are based on the spatial redundancy of the B image. 1229825 Calculate 1 ist 0 reference image, list 1 reference image and motion vector, and calculate the direct mode motion vector of the B image. 3 4. The method according to item 3 of the scope of patent application, where this step includes the following steps: If the neighboring blocks A, B, and C of the giant block to be coded represent different reference images, the reference image with the smallest index is selected as the reference image for each 1 ist. 3 5. If a patent is applied for The method according to item 33, wherein the step includes the following steps: if there are two or more reference images with the same index in the neighboring block of the giant block to be encoded, the reference image is selected as A reference image for each 1 ist. Method, wherein the step includes the following steps: if any of the neighboring blocks A, B, and C of the giant block to be encoded is in the 'in' mode, set its list 0 and list 1 motion vectors to 0; select a motion vector whose direction must be the same as the direction of the time position of the reference image of each list from a neighboring block; and, obtain the motion vector of each list through a median operation; or, if a neighboring When a block has two motion vectors in the same direction, only one of the two vectors is selected from the block, and the motion vector of each list is obtained by including the median operation of the selected motion vector. 37. If applying for a patent The method of scope item 33, wherein the step includes the following steps: if a valid reference image index for each 1 ist mode cannot be derived, set the list 0 reference image index and the list 1 reference image index to 0, and Set the motion vector of each list mode to 0. 第63頁Page 63