1290657 九、發明說明： 【發明所屬之技術領域】 本發明係有關一種為數位影像清除雜訊的方法及裝 置，特別是關於一種高效率的數位影像清除雜訊的方法。 【先前技術】 由於電腦的普及以及人們對於便利性的需求，使得數 位相機已漸漸地取代傳統相機的地位。對於數位相機而 言，所產生的影像品質乃是判斷數位相機優劣的重要依 據，因此各家廠商無不致力於影像品質的改善。在數位相 機中，若所拍攝未經處理的影像具有雜訊，該雜訊將嚴重 地影響經處理後影像的品質，這是因為在放大影像的色彩 值時，也同時將其上的雜訊放大，因此，在對所拍攝的影 像進行處理前，需先將其上的雜訊清除。 在已知的去雜訊方法中，西格瑪濾除（Sigma filter)法 是較佳的雜訊清除方式，第一圖a、b及c分別為士 (Bayer)彩色濾光片陣列所選取的綠色通道、紅色1、、务 色通道的5x5像素陣列1〇、12及14，在各陣歹彳^及现 及14中心位置的感興趣像素即為要進行雜 12 ^ 清除的傻 素，此法先估算每一鄰近像素的色彩值h 12 ^i^R〇 Μ Βι〜與各自感興趣像素的色彩值gc、Rc&b ^ c <間的差 值大小疋否大於一預設之臨界值TH ;接著，將 心位置像素色彩值Gc、Rc及Bc差值絕對值等於 ”中 界值ΤΉ的像素的色彩值與中心位置像素的色彩^於^ 乂逍Gc進行 1290657 加、、&平均’以得到一平均值取代中心位置像素的色彩值。 為更清楚說明，參照第二圖及第三圖A、B、c及D， 其中第二圖為西格瑪濾除法的流程圖，第三圖A、B、c 及D則為根據弟二圖流程圖所做之範例。第三圖a顯示一 綠色通道的5x5像素陣列，在步驟2〇中，估算所有鄰近 像素的色彩值在減去感興趣像素的色彩值後的絕對值，在 第三圖A中，中心位置（3,3)的感興趣像素的色彩值為20， 故將所有像素的色彩值減去20，各像素的色彩值減去2〇 後的絕對值如第三圖B所示。在步驟22中，將所有的絕 對值與預設之臨界值TH比較，’假設TH為10，然後步驟 24選取與感興趣像素的色彩值差值絕對值小於或等於臨 界值TH的鄰近像素，在此範例中，臨界值TH為1〇，因 此位置（1，1)、（1，5)、（2,4)、（3,5)、（4,2)及(4,4)的像素的色 彩值不予考慮，如第三圖C所示。在步驟26中，平均所 選取的鄰近像素及感興趣像素的色彩值而得到一平均值 21，最後步驟28，將平均值21取代感興趣像素的色彩值， 如第三圖D所示。同理紅色通道及藍色通道亦是根據上述 的步驟來達成雜訊的清除。 然而，由第一圖A、B及C中可以看出，在使用西格 瑪濾除法進行影像處理時，各色彩通道的像素陣列10、12 及14中所包含的像素數目是不同的，綠色通道中包括12 個鄰近像素，而紅色及藍色通道中包括8個鄰近像素，而 且每次均需計數絕對值大於及等於或小於臨界值TH的像 素數目，因而導致硬體執行時的複雜度’再者，判斷陣列 1290657 中所包含的像素數目也造成了時間的浪費。 因此，一種在為數位影像去雜訊時方便硬體運算的方 法乃為所冀。 【發明内容】 本發明的目的之一，在於提供一種為數位影像清除雜 訊的方法。 本發明的目的之一，另在於提供一種高效率的數位影 像去雜訊法。 根據本發明，一種為數位影像清除雜訊的方法，首先 在一感興趣像素的周圍選取一數目的鄰近像素；其次將每 一該鄰近像素的色彩值減去該感興趣像素的色彩值；接著 將鄰近像素的色彩值與該感興趣像素的色彩值差值絕對 值與一臨界值比較，若大於該預設之臨界值時，該鄰近像 素的色彩值便以該感興趣像素的色彩值取代；接著將所有 鄰近像素的色彩值平均得到一平均值來取代該感興趣像 素的色彩值。 【實施方式】 第四圖A、B及C分別為從一拜耳彩色濾光片陣列感 測器所產生的信號中選取的綠色通道、紅色通道及藍色通 道的5x5像素陣列30、32及34，在各陣列30、32及34 中心位置的像素即為要進行雜訊清除的感興趣像素。第五 圖顯示本發明之去雜訊裝置40。根據本發明的方法，在拜 1290657 色渡光片陣列感測器所產生的原始色彩值lA輸入裝 置^的像素陣列緩衝器42後，處理器45由原始色彩值 ^中分別選取綠色通道、紅色通道及藍色通道的像素陣列 ^、比及匕’例如第四圖所示的像素陣列^及^， :存^作記憶體44中。在對像素陣列&、Μ &進行 2 =時，先由記憶體46中讀取所要選取的鄰近像素 = 此數值由輸人SET1預先狀，較佳者為2的 人方，其中’若該感興趣像素的色彩為綠色時，所選取 ^近像素將在該感興趣像素的周圍排列成—菱形，若^ ::::素的色彩為紅色或藍色時，所選取的鄰近像素： ^感“趣像素的周圍排列成—正方形，如第四圖A』 ==?實:_，"為8。將咖 的色彩值及:及接H減t各自的感興趣像素 田 C接者再由記憶體48中讀取一 :值™，此數值由輸入咖預先設定，將TH與所有： 差值的絕對值比較，當任—鄰近像素的色彩值差值 中心位置像素的色彩值GC、R^BC取代，再= =中所有的鄰近像素分料行加總平均得到—平 ，各。自通道的中心位置像素的色彩值Gc、&及 ir Ε ^ ^ ^ ^ Gc^ Rc ^ Bc ^ # ^ # ^ 口口、，產生去除雜訊後的色彩值〇Α。 B、C為二楚說Γ述之f法’參照第六圖及第七圓心 、丁歹1其中第六圖為根據本發明之方法的流 1290657 程圖，第七圖A、B、C及D則為根據第六圖流程圖所做 之範例。在步驟50中，在感興趣像的周圍選取8個鄰近 像素，如第七圖A所提供之綠色通道的像素，接著步驟 52估算每一鄰近像素與感興趣像素的色彩值差值的絕對 值，如第七圖B所示，步驟54將每一鄰近像素與感興趣 像素的色彩值差值的絕對值與一預設之臨界值TH比較， 在此實施例中，臨界值TH為10，接下來步驟56將與感 興趣像素的色彩值差值的絕對值大於臨界值10的鄰近像 素的色彩值以感興趣像素的色彩值取代，在第七圖B中， 位置(2,4)、（3,5)、（4,2)及(4,4)的鄰近像素與中心位置（3,3) 的感趣像素的色彩值差值的絕對值大於10,因此這些像素 的色彩值以感興趣像素的色彩值20取代，如第七圖C所 示，再來步驟58平均所有鄰近像素的色彩值而得到一平 均值22,最後步驟59以該平均值22取代感興趣像素的色 彩值，如第七圖D所示。 由於本發明在任一色彩通道中所選取的鄰近像素的 數目均相同，因此不需再計數減去該感興趣像素後絕對值 等於或小於該臨界值的像素數目，故在做乘法運算時，將 使得硬體更容易處理。此外，本發明之雜訊清除方法可以 得到比習知西格瑪濾除方法更佳的效果。 第八圖、第九圖以及第十圖分別顯示以傳統的雜訊清 除方法、西格瑪濾除法以及本發明之方法模擬所得的曲線 圖，其中X軸表示所有像素灰階的平均值，Y軸表示與正 常色彩的標準差，標準差愈高則表示雜訊愈嚴重。在第八 1290657 =中，曲線ό〇為紅色色彩值在不同灰階平均值時的雜訊 払準差曲線，曲線62為藍色色彩值在不同灰階平均值時 的雜訊標準差曲線，曲線64為綠色色彩值在不同灰階平 均值時的雜訊標準差曲線，曲線66為亮度在不同灰階平 句值寺的雜5凡標準差曲線，在第九圖中，曲線7〇為紅色 色形值在不同灰階平均值時的雜訊標準差曲線，曲線72 為藍色色彩值在不同灰階平均值時的雜訊標準差曲線，曲 線74為綠色色彩值在不同灰階平均值時的雜訊標準差曲 線，曲，76為亮度在不同灰階平均值時的雜訊標準差曲 線’在第十圖中，曲線8G為紅色色彩值在不同灰階平 值時的雜訊標準差轉，㈣82為藍色色彩值在不同灰 P白平均值％的雜訊標準差曲線，曲線84為綠色色彩值 不同灰階平均值時的雜訊標準差曲線，曲線86為亮 不同灰階平均值時的雜訊標準差曲線。由第人圖^ ===比較，本發明之方法所得的結果優於傳統_ 濕π除方法，而與西格瑪濾除法所得的結果近似，甚至更 數攄1二傳統的雜訊清除方法的比· 數據，，、中色彩變化i〜13係由亮到暗的改變，… =紅色色彩值，G為綠色色彩值』為藍色色= :::四列欄位中的數值表示本發明之方法 與傳統的雜訊清除方法所得結果的差值， = :以百分比表示本發明之方法相對於傳統 = 度’攔位中的值為負值表示本發明所得的結果劣於傳^ 1290657 雜訊清除方法，由表一可以看出，具有正值的欄位遠多於 具有負值的欄位，本發明的方法遠優於傳統的雜訊清除方 法0 表一 色彩變化 1 2 3 4 5 6 7 8 9 10 11 12 13 平均値 γ的差値 0.21 -0.04 0 0.01 0.13 0.16 0.35 0.36 0.48 0.69 0.35 0.39 0.6 0.2838462 _差値 0.4 -0.06 -0.03 0.44 0.66 0.4 0.77 0.82 1.09 1.25 0.72 0.78 1.04 0.6369231 G的差値 0.17 -0.02 0.02 -0.05 0.05 0.13 0.34 0.35 0.42 0.8 0.39 0.36 0.54 0.2692308 B的差値 0.06 0.04 0.03 0.06 0.25 0.39 0.58 1 0.65 1.16 0.71 0.68 0.74 0.4884615 Y的差値 22% -6% 0% 1% 12% 6% 27% 20% 24% 31% 18% 19% 37% 16% 瞒差値 31% -6% -2% 24% 33% 13% 35% 28% 35% 33% 24% 27% 36% 24% G的差値 20% -3% 2% -4% 5% 5% 23% 19% 20% 33% 20% 17% 32% 14% B的差値 6% 4% 2% 4% 17% 12% 29% 37% 24% 36% 1S% 2S% 29% 19% 表一為本發明之方法與西格瑪濾除法的比較數據，同 樣地，色彩變化1〜13係由亮到暗的改變，γ表示亮度，R 為紅色色衫值，G為綠色色彩值，B為藍色色彩值，上方 的四列攔位中的數值表示本發明之方法所得的結果與西 格瑪濾除法所得結果的差值，而下方四列欄位則以百分比 表示本發明之方法相對於西格瑪濾除法的改良程度，攔位 中的值為負值表示本發明所得的結果劣於西格瑪濾除 去由表〜可以看出，具有正值的攔位仍多於具有負值的 攔位’本發明的方法略優於西袼瑪滤除方法。 表二 1290657 色彩變化 1 2 3 4 5 6 7 8 9 10 11 12 13 平均値 γ的差値 0.2 -0.02 0.03 -0.01 -0.01 0.08 -0.02 0.03 0.06 0.02 0.01 0.04 0 0.031538 啲差値 0.23 -0.06 0 0.29 0.15 0.05 -0.01 0.08 0.24 0.19 -0.05 0.17 0.12 0.107692 G的差値 0.2 -0.02 0.04 -0.04 0 0.06 -0.05 0.02 -0.02 -0.05 0 -0.05 -0.08 0.000769 B的差値 0.04 0.03 0.1 -0.02 -0.01 0.02 0.05 0.14 -0.06 -0.07 0.02 -0.02 0.01 • 0.017692 Y的差値 23% -3% 2% -1% -1% 3% -2% 2% 4% 1% 1% 2% 0% 2% 啲差値 22% -6% 0% 17% 10% 2% -1% 4% 10% 7% -2% 7% 6% 6% G的差値 23% -3% 3% -3% 0% 2% -5% 1% -1% -3% 0% -3% -7% 0% B的差値 4% 3% 7% -1% -1% 1% 3% 8% -3% -3% 1% -1% 1% 1% 以上對於本發明之較佳實施例所作的敘述係為闡明 之目的，而無意限定本發明精確地為所揭露的形式，基於 以上的教導或從本發明的實施例學習而作修改或變化是 鲁 可能的，實施例係為解說本發明的原理以及讓熟習該項技 術者以各種實施例利用本發明在實際應用上而選擇及敘 述，本發明的技術思想企圖由以下的申請專利範圍及其均 等來決定。 【圖式簡單說明】 對於熟習本技藝之人士而言，從以下所作的詳細敘述 配合伴隨的圖式，本發明將能夠更清楚地被瞭解，其上述鲁 及其他目的及優點將會變得更明顯，其中·· 第一圖A為由拜耳彩色濾光片陣列所選取的綠色渴 · 道的5x5像素陣列； 、 第一圖B為由拜耳彩色濾光片陣列所選取的紅色、南 的5x5像素陣列； — 第一圖C為由拜耳彩色濾光片陣列所選取的藍色通道 · 12 l29〇657 第二圖係西格瑪濾除法的流程圖； 第三圖A顯示一綠色通道的5x5像素陣列的範例； /第三HB顯示第三®JA巾各鄰近像素與感興趣像素的 色彩值差值的絕對值； 第三圖C顯示與感興趣像素的色彩值差值的絕對值小 於或等於預設之臨界值的鄰近像素； 第二圖D顯示感興趣像素在濾除雜訊後的色彩值； 第四圖A為本發明在一拜耳彩色濾光片陣列所選取 的綠色通道的5x5像素陣列及所選取的鄰近像素； 第四圖B為本發明在一拜耳彩色遽光片陣列所選取的 紅色通道的5x5像素陣列及所選取的鄰近像素； …第四圖C為本發明在一拜耳彩色遽光片陣列所選取的 藍色通道的5x5像素陣列及所選取的鄰近像素； 第五圖顯示本發明之去雜訊裝置的實施例； 第六圖為根據本發明之方法的流程圖； 第七圖A顯示一綠色通道的5χ5像素陣列的實施例； ，第顯示第六圖A巾各鄰近像素與感興趣像素的 色彩值差值的絕對值； ^第七HC顯示與感興趣像素色彩值差值的絕對值大於 預。又之界值的鄰近像素色彩值被感興趣像素的色彩值 取代後之示意圖； $七圖D顯示感興趣像素在濾、除雜訊後的色彩值； 第八圖係傳統雜訊清除方法所得的曲線圖； 第九圖係西格瑪濾除方法所得的A線®I ;以及 13 1290657 第十圖為本發明方法所得之曲線圖。 【主要元件符號說明】 綠色通道的5x5像素陣列 12 紅色通道的5x5像素陣列 14 Μ色通道的5x5像素陣列 2〇 估算每一鄰近像素與感興趣像素的色彩值差 值的絕對值 22 將每一鄰近像素與感興趣像素的色彩值差值 的絕對值與一預設之臨界值比較 24 選取與感興趣像素的色彩值差值的絕對值小 於或專於臨界值的鄰近像素 〜26 平均所選取鄰近像素及感興趣像素的色彩值 件到一平均值 9〇 、 以平均值取代感興趣像素的色彩值 3〇 綠色通道的5χ5像素陣列 32 紅色通道的5χ5像素陣列 34 藍色通道的5x5像素陣列 40 去除雜訊裝置 42 像素陣列緩衝器 44 操作記憶體 45 處理器 46 記憶體 1290657 50 在感興趣像素的周圍選取鄰近像素 52 估算每一鄰近像素與感興趣像素的色彩值差 值的絕對值 54 將每一鄰近像素與感興趣像素的色彩值的絕 對值與一預設之臨界值比較 56 將與感興趣像素的色彩值差值的絕對值大於 臨界值的鄰近像素的色彩值以感興趣像素的色彩值取代 58 平均所有鄰近像素的色彩值得到一平均值 59 以平均值取代感興趣像素的色彩值 60 紅色色彩值在不同灰階平均值時的雜訊標準 差曲線 62 藍色色彩值在不同灰階平均值時的雜訊標準 差曲線 64 綠色色彩值在不同灰階平均值時的雜訊標準 差曲線 66 亮度在不同灰階平均值時的雜訊標準差曲線 70 紅色色彩值在不同灰階平均值時的雜訊標準 差曲線 72 藍色色彩值在不同灰階平均值時的雜訊標準 差曲線 74 綠色色彩值在不同灰階平均值時的雜訊標準 差曲線 76 亮度在不同灰階平均值時的雜訊標準差曲線 80 紅色色彩值在不同灰階平均值時的雜訊標準 15 1290657 差曲線 82 藍色色彩值在不同灰階平均值時的雜訊標準 差曲線 84 綠色色彩值在不同灰階平均值時的雜訊標準 差曲線 86 亮度在不同灰階平均值時的雜訊標準差曲線1290657 IX. Description of the Invention: [Technical Field] The present invention relates to a method and apparatus for erasing noise for digital images, and more particularly to a method for removing noise from high-efficiency digital images. [Prior Art] Due to the popularity of computers and the demand for convenience, digital cameras have gradually replaced the status of traditional cameras. For digital cameras, the quality of the images produced is an important basis for judging the pros and cons of digital cameras. Therefore, all manufacturers are committed to improving image quality. In a digital camera, if the unprocessed image is taken with noise, the noise will seriously affect the quality of the processed image, because when the color value of the image is enlarged, the noise on the image is also simultaneously Zoom in, therefore, you need to clear the noise on the captured image before processing it. Among the known methods of denoising, the Sigma filter method is a better method of noise removal. The first figures a, b and c are the green colors selected by the Bayer color filter array. The channel, red 1, and 5x5 pixel arrays of the color channel 1〇, 12, and 14 are the pixels of interest in each of the arrays and the center of the 14th position. First, it is estimated that the color value h 12 ^i^R〇Μ Βι~ of each neighboring pixel and the color value gc, Rc & b ^ c < of the respective pixel of interest are not greater than a predetermined threshold. TH; Next, the absolute value of the difference between the heart position pixel color values Gc, Rc, and Bc is equal to the color value of the pixel of the middle boundary value 与 and the color of the pixel of the center position ^^ 乂逍Gc is 1290657 plus, & average 'To replace the color value of the central position pixel with an average value. For more clarity, refer to the second figure and the third figure A, B, c and D, wherein the second picture is a flow chart of the sigma filtering method, the third figure A, B, c, and D are examples based on the flowchart of the second diagram. The third diagram a shows a green pass. The 5x5 pixel array, in step 2, estimates the absolute value of the color values of all neighboring pixels after subtracting the color values of the pixel of interest. In the third graph A, the pixel of interest at the center position (3, 3) The color value is 20, so the color value of all pixels is subtracted by 20, and the absolute value of each pixel's color value minus 2〇 is shown in Figure B. In step 22, all absolute values are compared with Let the threshold TH be compared, 'assuming TH is 10, then step 24 selects neighboring pixels whose absolute value of the color value difference of the pixel of interest is less than or equal to the threshold TH, in this example, the threshold TH is 1〇, Therefore, the color values of the pixels of the positions (1, 1), (1, 5), (2, 4), (3, 5), (4, 2), and (4, 4) are not considered, as shown in the third figure. In step 26, the color values of the selected neighboring pixels and the pixel of interest are averaged to obtain an average value 21, and finally, step 28, the average value 21 is substituted for the color value of the pixel of interest, as shown in the third figure D. As shown in the figure, the red channel and the blue channel are also cleared according to the above steps. However, by the first figure A It can be seen from B and C that the number of pixels included in the pixel arrays 10, 12 and 14 of each color channel is different when the image processing is performed by the sigma filtering method, and the green channel includes 12 adjacent pixels, and the red color And the blue channel includes 8 neighboring pixels, and each time it is necessary to count the number of pixels whose absolute value is greater than or equal to or less than the threshold TH, thus causing the complexity of the hardware execution 'again, judging the inclusion in the array 1290657 The number of pixels also causes a waste of time. Therefore, a method for facilitating hardware operations when removing noise for digital images is a problem. SUMMARY OF THE INVENTION One object of the present invention is to provide a method for clearing noise for a digital image. One of the objects of the present invention is to provide a highly efficient digital image denoising method. According to the present invention, a method for removing noise from a digital image first selects a number of adjacent pixels around a pixel of interest; secondly, subtracts the color value of each of the adjacent pixels from the color value of the pixel of interest; Comparing the color value of the adjacent pixel with the absolute value of the color value difference of the pixel of interest and a threshold value, if greater than the preset threshold value, the color value of the neighboring pixel is replaced by the color value of the pixel of interest The color values of all neighboring pixels are then averaged to obtain an average value to replace the color value of the pixel of interest. [Embodiment] The fourth figure A, B and C are respectively 5x5 pixel arrays 30, 32 and 34 of the green channel, the red channel and the blue channel selected from the signals generated by a Bayer color filter array sensor. The pixels at the center of each of arrays 30, 32, and 34 are the pixels of interest to be cleared by noise. The fifth figure shows the de-noising device 40 of the present invention. According to the method of the present invention, after the original color value 1A generated by the 1290657 color channel array sensor is input to the pixel array buffer 42 of the device, the processor 45 selects the green channel and the red color from the original color values ^, respectively. The pixel arrays of the channels and the blue channels, the ratios, and the 像素', for example, the pixel arrays ^ and ^ shown in the fourth figure are stored in the memory 44. When 2 = is performed on the pixel array &, Μ &, the neighboring pixel to be selected is first read from the memory 46 = the value is pre-shaped by the input SET1, preferably the person of 2, where 'if When the color of the pixel of interest is green, the selected near pixels will be arranged in a diamond shape around the pixel of interest, and if the color of the ::::: element is red or blue, the selected neighboring pixels: ^Feeling "the surrounding pixels are arranged in a square - as in the fourth picture A" ==? Real: _, " is 8. The color value of the coffee and: and H minus t each of the pixel of interest Then, a value of TM is read from the memory 48. This value is preset by the input coffee, and TH is compared with all: the absolute value of the difference, and the color value of the pixel at the center of the difference between the color values of the adjacent pixels GC, R^BC is replaced, and then all the adjacent pixel sub-rows are averaged to obtain - flat, each. The color values of the pixels from the center of the channel are Gc, & and ir Ε ^ ^ ^ ^ Gc^ Rc ^ Bc ^ # ^ # ^ 口 口 , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , And the seventh center, the first one, wherein the sixth figure is the flow 1290657 process according to the method of the present invention, and the seventh figure A, B, C and D are the examples according to the flow chart of the sixth figure. The eight adjacent pixels are selected around the image of interest, such as the pixels of the green channel provided in FIG. 7A, and then step 52 is used to estimate the absolute value of the difference between the color values of each adjacent pixel and the pixel of interest, such as As shown in FIG. 24, step 54 compares the absolute value of the difference between the color values of each neighboring pixel and the pixel of interest with a predetermined threshold TH. In this embodiment, the threshold TH is 10, and the next step is 56. The color value of the neighboring pixel whose absolute value of the color value difference of the pixel of interest is greater than the threshold value 10 is replaced by the color value of the pixel of interest. In the seventh diagram B, the position (2, 4), (3, 5), (4, 2) and (4, 4) the adjacent pixel and the central position (3, 3) of the pixel value difference value of the absolute value is greater than 10, so the color value of these pixels in the pixel of interest Substituting the color value 20, as shown in Figure 7, and then step 58 averages the color values of all neighboring pixels. An average value 22 is obtained, and the final step 59 replaces the color value of the pixel of interest with the average value 22, as shown in the seventh figure D. Since the number of adjacent pixels selected in any color channel is the same, the present invention does not The number of pixels whose absolute value is equal to or less than the critical value after subtracting the pixel of interest needs to be counted again, so that the hardware is more easily processed when multiplication is performed. In addition, the noise clearing method of the present invention can be obtained. The sigma filtering method has better effects. The eighth, ninth and tenth graphs respectively show the graphs obtained by the conventional noise scavenging method, the sigma filtering method and the method of the present invention, wherein the X axis represents all The average of the grayscale of the pixel, the Y-axis represents the standard deviation from the normal color, and the higher the standard deviation, the more severe the noise. In the eighth 1290657 =, the curve ό〇 is the noise 払 standard curve of the red color value at different gray level averages, and the curve 62 is the noise standard deviation curve of the blue color value at different gray level averages, Curve 64 is the noise standard deviation curve of the green color value at different gray level averages, and curve 66 is the standard deviation curve of the brightness in the different gray level flat sentence value temple. In the ninth figure, the curve 7〇 is The noise standard deviation curve of the red color shape value at different gray level average values, the curve 72 is the noise standard deviation curve of the blue color value at different gray level average values, and the curve 74 is the green color value at different gray level averages. The noise standard deviation curve of the value, the curve, 76 is the noise standard deviation curve of the brightness at different gray level averages. In the tenth figure, the curve 8G is the noise of the red color value at different gray level values. Standard deviation, (4) 82 is the noise standard deviation curve of blue color value in different gray P white average value, curve 84 is the noise standard deviation curve when the green color value is different gray level average, curve 86 is bright gray The standard deviation curve of noise at the average of the order. Compared with the traditional _ wet π removal method, the results obtained by the method of the present invention are similar to those obtained by the sigma filter method, and even more than the conventional noise removal method. · Data,,, medium color change i~13 is a change from light to dark, ... = red color value, G is green color value 』 is blue color = ::: the value in the four columns indicates the method of the present invention The difference from the result obtained by the conventional noise removal method, =: the percentage value of the method of the present invention relative to the conventional = degree 'blocking' is expressed as a percentage, indicating that the result obtained by the present invention is inferior to the transmission of 1290657 noise removal. The method, as can be seen from Table 1, is that the field with positive values is far more than the field with negative values, and the method of the present invention is far superior to the conventional method for removing noise. Table 1 Color Change 1 2 3 4 5 6 7 8 9 10 11 12 13 The difference between the average 値γ is 0.21 -0.04 0 0.01 0.13 0.16 0.35 0.36 0.48 0.69 0.35 0.39 0.6 0.2838462 _Evaluation 0.4 -0.06 -0.03 0.44 0.66 0.4 0.77 0.82 1.09 1.25 0.72 0.78 1.04 0.6369231 G 0.17 -0.02 0.02 -0.05 0.05 0.13 0.34 0.35 0 .42 0.8 0.39 0.36 0.54 0.2692308 B's difference 0.06 0.04 0.03 0.06 0.25 0.39 0.58 1 0.65 1.16 0.71 0.68 0.74 0.4884615 Y difference of 22% -6% 0% 1% 12% 6% 27% 20% 24% 31% 18% 19% 37% 16% 瞒 difference 31% -6% -2% 24% 33% 13% 35% 28% 35% 33% 24% 27% 36% 24% G difference 20% -3% 2% -4% 5% 5% 23% 19% 20% 33% 20% 17% 32% 14% B 6% 6% 4% 2% 4% 17% 12% 29% 37% 24% 36% 1S % 2S% 29% 19% Table 1 is the comparison data between the method of the present invention and the sigma filtering method. Similarly, the color changes 1 to 13 are changed from light to dark, γ is the brightness, and R is the red color value, G For the green color value, B is the blue color value. The value in the upper four columns of the block indicates the difference between the result obtained by the method of the present invention and the result obtained by the sigma filter method, and the lower four columns are expressed as a percentage. The degree of improvement of the method of the invention relative to the sigma filter method, the value of the negative value in the block indicates that the result obtained by the present invention is inferior to the sigma filter removed. It can be seen that the block with positive values still has more negative values. Interceptor's invention France is slightly better than the West eligible Mary filtering method. Table 2 1290657 Color change 1 2 3 4 5 6 7 8 9 10 11 12 13 The difference between the average 値 値 0.2 -0.02 0.03 -0.01 -0.01 0.08 -0.02 0.03 0.06 0.02 0.01 0.04 0 0.031538 啲 値 値 0.23 -0.06 0 0.29 0.15 0.05 -0.01 0.08 0.24 0.19 -0.05 0.17 0.12 0.107692 G difference of 0.2 -0.02 0.04 -0.04 0 0.06 -0.05 0.02 -0.02 -0.05 0 -0.05 -0.08 0.000769 B difference 0.04 0.03 0.1 -0.02 -0.01 0.02 0.05 0.14 -0.06 -0.07 0.02 -0.02 0.01 • 0.017692 The difference between Y and 23% -3% 2% -1% -1% 3% -2% 2% 4% 1% 1% 2% 0% 2% 啲 値22% -6% 0% 17% 10% 2% -1% 4% 10% 7% -2% 7% 6% 6% G difference of 23% -3% 3% -3% 0% 2% - 5% 1% -1% -3% 0% -3% -7% 0% B difference of 4% 3% 7% -1% -1% 1% 3% 8% -3% -3% 1% -1% 1% 1% The above description of the preferred embodiments of the present invention is for illustrative purposes, and is not intended to limit the invention to the precise form disclosed, or based on the teachings of the invention. It is possible to make modifications or variations, and the embodiments are intended to illustrate the principles of the invention and to enable those skilled in the art to In practice of the present invention is selected and the above classification, the technical idea of the present invention is an attempt by the following claims and their like are determined. BRIEF DESCRIPTION OF THE DRAWINGS The present invention will be more clearly understood from the following detailed description taken in conjunction with the accompanying drawings. Obviously, the first picture A is a 5x5 pixel array of green thirsty channels selected by the Bayer color filter array; the first picture B is the red and south 5x5 selected by the Bayer color filter array. Pixel array; - Figure C is a blue channel selected by the Bayer color filter array. 12 l29 〇 657 second flowchart sigma filtering method; third panel A shows a green channel 5x5 pixel array The third example shows the absolute value of the difference between the color values of the neighboring pixels and the pixel of interest of the third ® JA towel; the third graph C shows that the absolute value of the difference between the color values of the pixel of interest is less than or equal to the pre- The neighboring pixels of the critical value are set; the second figure D shows the color values of the pixel of interest after filtering the noise; the fourth figure A is the 5x5 pixel array of the green channel selected by the Bayer color filter array of the present invention. And The adjacent pixel is taken; the fourth picture B is a 5x5 pixel array of the red channel selected by the Bayer color enamel array in the present invention and the selected adjacent pixel; the fourth picture C is a Bayer color ray in the present invention. a 5x5 pixel array of blue channels selected by the slice array and selected neighboring pixels; a fifth diagram showing an embodiment of the de-noising device of the present invention; and a sixth diagram showing a flow chart according to the method of the present invention; A shows an embodiment of a 5 χ 5 pixel array of a green channel; the first display shows the absolute value of the color value difference between each adjacent pixel and the pixel of interest of the sixth picture A; ^ the difference between the seventh HC display and the color value of the pixel of interest The absolute value is greater than the pre-. The figure of the neighboring pixel color value of the boundary value is replaced by the color value of the pixel of interest; $7D shows the color value of the pixel of interest after filtering and removing noise; The eighth figure is obtained by the traditional noise clearing method. The ninth graph is the A-line®I obtained by the sigma filtration method; and the 13th 1290657 is the graph obtained by the method of the present invention. [Main component symbol description] 5x5 pixel array of green channel 12 5x5 pixel array of red channel 14 5x5 pixel array of color channel 2〇 Estimate the absolute value of the difference between the color value of each adjacent pixel and the pixel of interest 22 Comparing the absolute value of the difference between the color value of the neighboring pixel and the pixel of interest with a preset threshold value 24 selecting the neighboring pixel whose absolute value of the color value difference with the pixel of interest is smaller than or specific to the threshold value Color values of adjacent pixels and pixels of interest to an average value of 9〇, replacing the color value of the pixel of interest with an average value 3〇 5χ5 pixel array of green channel 32 5χ5 pixel array of red channel 34 5x5 pixel array of blue channel 40 De-noising device 42 Pixel array buffer 44 Operating memory 45 Processor 46 Memory 1290657 50 Selecting neighboring pixels 52 around the pixel of interest Estimating the absolute value of the difference in color value between each neighboring pixel and the pixel of interest 54 Comparing the absolute value of the color value of each neighboring pixel with the pixel of interest with a predetermined threshold value 56 will be compared with the image of interest The color value of the adjacent pixel whose absolute value of the color value difference is greater than the critical value is replaced by the color value of the pixel of interest. 58 Average the color value of all neighboring pixels to obtain an average value 59. The color value of the pixel of interest is replaced by the average value. The noise standard deviation curve of the color value at different gray level averages 62 The noise standard deviation curve of the blue color value at different gray level averages 64 The noise standard deviation curve of the green color values at different gray level averages 66 Noise standard deviation curve of brightness at different gray level averages 70 Noise standard deviation curve of red color value at different gray level averages 72 Noise standard deviation curve of blue color values at different gray level averages 74 Green color value The noise standard deviation curve of different gray level averages 76 The noise standard deviation curve of brightness at different gray level averages 80 Red color value The noise standard of different gray level averages 15 1290657 Poor Curve 82: The standard deviation of the noise of the blue color value at different gray level averages 84 The noise standard deviation curve of the green color value at different gray level averages 86 Brightness Noise standard deviation curve at different gray level averages
1616