CN103986930A - Registration picture image compression method suitable for small-capacity information carrier - Google Patents
Registration picture image compression method suitable for small-capacity information carrier Download PDFInfo
- Publication number
- CN103986930A CN103986930A CN201410242515.2A CN201410242515A CN103986930A CN 103986930 A CN103986930 A CN 103986930A CN 201410242515 A CN201410242515 A CN 201410242515A CN 103986930 A CN103986930 A CN 103986930A
- Authority
- CN
- China
- Prior art keywords
- image
- registration
- data
- human face
- coding
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Landscapes
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention relates to a registration picture image compression method suitable for a small-capacity information carrier. The registration picture image compression method includes the steps of information collecting, image preprocessing, frequency domain converting, quantizing and entropy coding, and cod stream compressing. The registration picture image compression method is characterized in that targeted preprocessing is conducted on a registration picture image, the image is divided into a human face part, a clothing part and a background part, and different processing methods are adopted in the different parts. In the coding process, the resolution, the image type and the coding parameters of the image are set in advance, and code stream control information and image description information in code streams of image coding are little; in the decoding process, error detection can be conducted on partitioned data of the human face area, and image quality losses generated by data error codes of image blocks are reduced.
Description
Technical field
The present invention relates to a kind of Image Compression, be especially applicable to capacity compared with realizing registration on vectorette according to method for compressing image.
Background technology
In the Image Compression generally using at present, for cromogram and gray-scale map, conventional stationary image compression coding standard has two kinds:
One is the i.e. abbreviation of " continuous tone rest image digital compression coding " international standard of JPEG:JPEG, is the common method of image compression.
JPEG is the Standard of image compression taking piecemeal and dct transform (discrete cosine transform) as core, uses very generally in common image compression, but still has many defects while being applied to registration according to image Compression; Be mainly manifested in:
(1) in file format, contain a lot of description information of images and code stream control information, when compressed image hour, these redundant informations are very large on reducing the impact of compression efficiency;
(2) data blocks all in image is adopted to identical compress mode, the important information in cannot shining registration: face is laid special stress on protecting;
(3) description information of image in file format and code stream control information antijamming capability are too poor, the serious consequence that even there will be data cannot decode completely in the time that these information are interfered.
Its two for JPEG2000:JPEG2000 be the international standard of a novel stationary image compression coding.
JPEG2000 has mainly adopted DWT conversion (wavelet transformation) to substitute dct transform, has improved code efficiency, also has a following defect but be applied to when registration is processed according to image:
(1) be first to have reduced compression efficiency as a kind of data of the requisite file header of file format of standard;
(2) secondly it adopts demonstration decoding gradually and the gradable embedded bitstream design of SNR to be more suitable for Internet Transmission, while Showing Picture on webpage, utilize and show that gradually decoding can provide better user to experience, but there is no need for the data that deposit multidimensional code in.
In addition, Quick Response Code, three-dimensional code, RFID etc. are the label forms of current extensive use, but as label, its information capacity is little, and view data generally needs certain data capacity.When needs when document image, often exist the problem of off-capacity on this class label.Conventional images compression standard not only defines the coding method of view data, has also formulated the form of file simultaneously, in image file, except the data for Description Image information, is also useful on the control information of Description Image form and coding parameter.Based on the image file of these Image Coding standards generations, be not suitable for the particular requirement of the low capacity such as Quick Response Code, RFID information carrier.
Therefore, how on low capacity information carrier, to shine image according to the registration of specific needs record, and seek a kind ofly to there is the image compression encoding method that compression efficiency is high, antijamming capability is strong, picture quality is good and there is certain practical significance, be also the technical task that relevant industries need to solve.
Summary of the invention
Object of the present invention: aim to provide a kind of method for compressing image of more simplifying packed data that has, not only improve the antijamming capability of image compression data, and can improve the picture quality after packed data decompress(ion).The present invention is achieved through the following technical solutions foregoing invention object
This registration that is applicable to low capacity information carrier the present invention relates to is according to method for compressing image, comprise information gathering, image preliminary treatment, frequency domain conversion, quantification and entropy coding, compressed bit stream, it is characterized in that: registration is carried out to specific aim preliminary treatment according to image, image is divided into face, clothing, the several parts of background, different piece is adopted to different processing methods; When coding, the resolution to image, image type, coding parameter are made an appointment, and do not contain any description information of image in image compression code stream, only need a small amount of code stream control information.
Concrete treatment step is as follows:
A. in video preview window, predetermined registration, according to the profile of image, is taken pictures by design profile;
B. registration is shone into the preliminary treatment of row image, according to the feature of image in picture, by adopting edge detection algorithm or judge rgb value or the method for gray value, picture is shone to background according to personage's face, registration and clothing is divided into 2-3 part;
C. registration is carried out to dct transform according to image, the human face region in image is carried out to small data piece and cut apart, and fixed block number; Long data block is carried out in all the other regions to be cut apart; For the image of prior agreement resolution, the detection of blocks of data mistake while for the benefit of decoding, the number of blocks of human face region should be fixed, and therefore the image block quantity in other region is also fixed;
D. the data after frequency domain conversion are quantized and entropy coding;
E. in image code stream, data are divided into human face region image block data and other area image blocks of data, human face region image block data is placed on the foremost of code stream, and human face region end-of-data mark is in the end set, utilize end of block flag and human face region end mark, carry out data block automatic error detection.
While registration image being divided into background, face, clothing three part according to the dividing method described in step B, background is set as to single value of color or gray value, most of region of clothing is made as to single color-values or gray value.
While registration image being divided into background, personage's two parts according to the dividing method described in step B, background is set as to single value of color or gray value.
Before the compressed bit stream of human face region being come according to the compressed bit stream arrangement mode described in step D, after the compressed bit stream in other regions comes.
Frequency domain conversion adopts dct transform; Entropy coding adopts Huffman coding, predicted difference Coded and run-length encoding; The long data block of cutting apart is 8X8 pixel, and small data piece is 4X4 pixel.Described frequency domain conversion can also adopt Karhunen-Loeve transformation, Walsh-Hadanjard Transform or wavelet transformation; Entropy coding can also adopt arithmetic coding; The size that is used for the data block of cutting apart can also be 16X16 pixel.
This registration proposing according to above technical scheme, according to method for compressing image, not only provides technical support according to the registration of specific needs record according to image for solving on low capacity information carrier, and has had the packed data of more simplifying.Not only improve the antijamming capability of image compression data, and can improve the picture quality after packed data decompress(ion).
Brief description of the drawings
Fig. 1 is image compression encoding schematic flow sheet;
Fig. 2 .1 ?Fig. 2 .3 be contour shape design example schematic diagram;
Fig. 3 is for registration is according to sample graph one;
Fig. 4 is for registration is according to sample graph two;
Fig. 5 is the histogram before preliminary treatment;
Fig. 6 is pretreated histogram;
Fig. 7 is the divided example schematic of samples pictures;
Fig. 8 is for registration is according to former figure;
Fig. 9 is the image that adopts commonsense method decompress(ion);
Figure 10 is the image adopting after the inventive method decompress(ion);
Data code flow after Figure 11 coding schematic diagram that puts in order;
Figure 12 is packed data decoding process figure.
In figure: 1 ?image background portion 2 ?image portrait portion.
Embodiment
This registration that is applicable to low capacity information carrier is according to method for compressing image, comprise information gathering, image preliminary treatment, frequency domain conversion, quantification and entropy coding, produce compressed bit stream, it is characterized in that: registration is carried out to specific aim preliminary treatment according to image, image is divided into face, clothing, the several parts of background, different piece is adopted to different processing methods; When coding, the resolution to image, image type, coding parameter are made an appointment, and code stream control information and description information of image in the code stream of Image Coding are little.While reaching decoding, can carry out error detection occurs to the block data of human face region, reduce the image quality loss that image block data error code produces.
For the feature of registration photograph, image is made an appointment according to coloured typies, image resolution ratio, sampling precision, coding parameter etc., therefore in image compression code stream without any need for description information of image, the only a small amount of code stream control information of needs.
Illustrate: in this specific embodiment, frequency domain conversion adopts dct transform; Entropy coding adopts Huffman coding, predicted difference Coded and run-length encoding; The long data block of cutting apart is 8X8 pixel, and small data piece is 4X4 pixel.But in the present invention, frequency domain conversion also can adopt additive method as Karhunen-Loeve transformation, Walsh-Hadanjard Transform and wavelet transformation etc.; Entropy coding also can adopt additive method as arithmetic coding etc.; The size that is used for the data block of cutting apart can be 4X4,8X8 or 16X16 pixel etc.
Concrete treatment step is as follows:
A. in video preview window, predetermined registration, according to the profile of image, is taken pictures by design profile;
B. registration is shone into the preliminary treatment of row image, according to the feature of image in picture, by adopting edge detection algorithm or judge the method for rgb value, picture is shone to background according to personage's face, registration and clothing is divided into 2-3 part;
C. registration is carried out to dct transform according to image, the human face region in image is carried out to 4X4 small data piece and cut apart also fixed block number; 8X8 long data block is carried out in all the other regions and cut apart also fixed block number; For the image of prior agreement resolution, the detection of blocks of data mistake while for the benefit of decoding, the number of blocks of human face region should be fixed, and therefore the image block quantity in other region is also fixed;
D. the data after frequency domain conversion are quantized and entropy coding;
E. in the image code stream after entropy coding, data are divided into human face region view data and other area image data.Human face region view data is placed on the foremost of code stream, and human face region end-of-data mark is in the end set, and an end of block flag is set after the data block in each region; Automatic error detection when end of block flag and human face region end mark can help to decode.
While registration image being divided into background, face, clothing three part according to the dividing method described in step B, background is set as to single value of color or gray value, most of region of clothing is made as to single color-values or gray value.
While registration image being divided into background, personage's two parts according to the dividing method described in step B, background is set as to single value of color or gray value.
Further set forth the present invention below in conjunction with above technical scheme.
This registration, according to method for compressing image, comprises information gathering, image preliminary treatment, dct transform, quantification and entropy coding, compressed bit stream according to the image compression encoding flow process described in Fig. 1.
In the time carrying out IMAQ, in the time carrying out IMAQ, in video preview window, the profile of predefine face (or personage), takes pictures by profile.This photo collecting according to ad hoc approach is convenient to preliminary treatment and the block encoding of image, makes successive image encoding efficiency better.
Accompanying drawing 2.1 ?2.3 have provided three kinds of preview pane (shape of profile, the design of size is not only limited to listed three kinds).As accompanying drawing 2.1 ?shown in three width figure as shown in 2.3, preview pane when rectangle housing represents to carry out image information collecting, frame inner region represents the profile for locking personage, starts to gather pictorial information in the time that personage is in profile.Accompanying drawing 2.1 ?2.3 what provide is that image is divided into two parts: top is image background portion 1, and bottom is image portrait portion 2.
In the time carrying out image preliminary treatment, consider that face registration is according to an important difference of image and common picture, the main information of image is the face that information spinner that people pay close attention to will concentrate on personage, and the background of registration photograph, personage's clothing are than minor information.The common registration that human eye is seen according in background be monochromatic, but in fact the rgb value (or gray value) of the each pixel in background is slightly variant; In clothing, the difference of the rgb value of pixel (or gray value) and saltus step are more obvious.Can produce many nonzero values if directly carry out dct transform, reduce compression efficiency.Therefore need background and clothing part to carry out corresponding preliminary treatment.
Scheme 1 is that image is divided into three parts:
(a) by edge detection algorithm or judge that image is divided into face by the method for rgb value (or gray value), background, clothing three parts.
(b) all pixels in background piece are all set to single value of color or gray value, for example, can be set to for cromogram the sky blue that rgb value is (240,255,255).
(c) clothing part is carried out to picture smooth treatment, to reduce the saltus step of image color value or gray value, reduce high fdrequency component.The method of smoothing processing can adopt the filtering of normalization piece, gaussian filtering, medium filtering scheduling algorithm.
Scheme 2 is that image is divided into two parts (see Fig. 2 .1 ?2.3):
(a) by edge detection algorithm or judge that the method for rgb value (or gray value) is divided into personage by image, background two parts.
(b) all pixels in background piece are all set to single value of color or gray value.
Background and clothing are carried out to pretreated effect displaying:
(1) the unification processing of the value of color of background parts (or gray value):
Shine image as example taking the registration in Fig. 3, the pixel in background is all unified into single rgb value, RGB=(240,255,255), has only been left DC component after dct transform, and the data block after conversion is extremely simplified.Only show G component result after treatment herein, R, B component, and the Y being changed into by RGB, U, tri-components of V have identical result.Table 1 is the G component value of registration according to upper left corner 8X8 block of pixels; Table 2 is that the data in his-and-hers watches 1 are carried out the dct transform result before preliminary treatment; Table 3 is pretreated dct transform results.
Table 1.G component value
Dct transform result before the preliminary treatment of table 2. data
The pretreated dct transform result of table 3. data
(2) smoothing processing of clothing part:
It is that clothing can show personage's aspectual character (as personage's fat or thin degree) that clothing part is not adopted to the reason of establishing single value of color (or gray value), thus can not directly carry out unification processing, but adopt smoothing processing.Shine image as example taking the registration in Fig. 4, clothing part is carried out to smoothing processing (what apply in this example is Gassian low-pass filter algorithm), after dct transform, the quantity of numerical value point bigger than normal has tailed off, and therefore after DCT piece quantizes, stays for entropy coded data and also can reduce.
Only show G component result after treatment herein, R, B component, and the Y being changed into by RGB, U, tri-components of V have identical result.Table 4 is registration G component values according to lower left corner 8X8 block of pixels; Table 5 is results that the data in his-and-hers watches 1 are carried out the dct transform before preliminary treatment and rounded; Table 6 is pretreated dct transform the result that rounds.In order more clearly to describe effect, the data in his-and-hers watches 5 and table 6 have been carried out statistics with histogram, and as shown in Figure 5 and Figure 6: abscissa represents 0～255 data sectional, ordinate represents the number that this hop count value is corresponding.
Table 4 is registration G component values according to lower left corner 8X8 block of pixels;
Dct transform before the preliminary treatment of table 5. data the result rounding
The pretreated dct transform of table 6. data the result rounding
Described in C, carrying out, before dct transform, the zones of different in image being carried out to the piecemeal of different pieces of information block size.
During due to collection image, people information is to gather according to the profile of making an appointment, so the position size of human face region is known and can fixes in advance in the image gathering.Registration taking Fig. 3 intermediate-resolution as 48X72 is according to being example, and it need to carry out region that small data piece cuts apart as shown in Figure 7.Fig. 7 is the data block that samples pictures has been divided into 54 8X8, and face characteristic major part all concentrates on the region of Fig. 7 central authorities grey, in order to ensure the picture quality in this region, the image block in this region is divided into the data block of smaller szie.
Still shine as example taking the registration in Fig. 3, common method of partition (image is divided into the data block of formed objects) is shone and compresses and decode this registration with the identical compression quality for selectivity method of partition (compression quality Q=70% in this example) in the present invention.Observe the image that obtains of decoding, adopt the compression method of selectivity piecemeal to make to decode after the picture quality of human face region had very large improvement, more close with the feature of original image.
Fig. 8, Fig. 9, what Figure 10 showed is respectively the former figure of BMP form, the image after the image after common dividing method compression coding and employing selectivity dividing method compression coding:
(4) dct transform, quantizes and entropy coding
Illustrate: dct transform, quantification and entropy are encoded not within the scope of claim of the present invention, are only briefly described:
(a) dct transform:
Taking the data block of 8X8 as example, the mathematical definition of its direct transform (FDCT), inverse transformation (IDCT) is provided by following equation:
FDCT:
IDCT：
Wherein:
(b) quantize: quantification can reduce amount of compressed data.
(c) entropy coding:
The key step of entropy coding is that DC component is carried out to difference prediction coding, and AC portion is carried out to run-length encoding, finally unified Huffman coding or the arithmetic coding of carrying out.Can the DC component in small data piece region not carried out to difference prediction coding in the present invention, to avoid error code to bring linksystem impact to small data piece region.
(d) putting in order and the embedding of control information of blocks of data:
Before human face region being arranged in this method, other area arrangement in the back.This sort method can avoid the error propagation of other regions generations in human face region.After each data block, there is end of block flag, after human face region, have human face region end mark, for distinguishing human face region and other regions.
Still taking the sample of Fig. 3 as example, human face region is divided into the data block of 4X4, other Region Segmentation are the data block of 8X8, one of arrangement mode of its blocks of data is as shown in figure 11: blue round dot is the starting point that data block is arranged, it is worthy of note: although emphasize the data of human face region to come before other regions in the present invention, the sequence of each intra-zone data block is not unique.
2. the image decoding stage:
Automatic error detection is decoded for the first time to packed data:
Data block is the general random length coding that adopts carry out entropy coding after dct transform time.Random length coding can greatly improve code efficiency, but therefore also needs to introduce control information: end of block flag is come the position of the each data block of mark.
Bit stream data produces error code and is divided into two kinds of situations:
(1) code stream in data block is by error code;
(2) control information is by error code.
Wherein the error code of control information in two kinds of situation: a kind of is after control information is originally disturbed, to have become data to sneak in data code flow; Another kind is that data code flow is originally disturbed and has produced the bit identical with control information and represent afterwards, has become a pseudo-control information.
In data block, the error code of code stream only can affect the decoding of this data block, but the error code of control information can have a strong impact on the decoding of subsequent data blocks.Because the data volume that control information accounts in whole compressed bit stream is little, therefore general impaired control information negligible amounts.This algorithm is for solving the detection that a small amount of control information is gone wrong.
Illustrate taking the sample of Fig. 3 as example, algorithm is for detection of the situation of a block end information error code of human face region.For the image of other resolution, algorithm can be by that analogy.If in this object lesson, important area small data piece block number be 64.The bit of end of block flag is expressed as: 0000; Human face region end mark is expressed as: 11111111.Error detection algorithm while decoding for key data is divided into following steps:
(a) according to block end identifier, data block is read one by one, block end number is counted, check a byte after block end simultaneously, if there are 81 in this byte, check whether end of block flag is now the 63rd, 64 or 65, if so, forwards step (c) to; Otherwise forward (b) to;
(b) again check a byte after end of block flag, if there are 71 in this byte, check whether end of block flag is now the 63rd, and if so, 64 or 65 be 1 by 0 bit position in byte, forwards step (c) to; Otherwise no longer find, forward (e) to
(c) determined the length and location of significant data.If the number of end of block flag is 64, directly decode.Due to the distribution comparing class of low-and high-frequency component in the each fritter of human face region seemingly, therefore the data flow length of each data block is more or less the same.Add up the data flow length l of each data block
1～l
63; If EOB number is 63, max (l
n) corresponding n number piece is to have two data blocks to be combined into, and can determine other not data block location of error code, forwards (d) to.If the number of end of block flag is 65, min (l
n+ l
n+1) corresponding No. n and the end of block flag of n+1 number interblock be that error code causes, and forwards (d) to.
(d) when end of block flag number is 63, two data blocks of mark error code; When end of block flag number is 65, a data block of mark error code; Forward (f) to.
(e) compressed bit stream is started to order decoding.
(f) correct data block is decoded, wrong data block is given up, and adopt the various algorithms of error concealing to repair decoded data.
(2) decoding: decode procedure method used, in this patent requires, is not only briefly described.Mainly contain following steps: entropy decoding, inverse quantization, DCT inverse transformation, YUV component is converted into RGB component, obtains decoded image.
Above are only part preferred embodiment of the present invention, the present invention is not limited in the content of embodiment.To those skilled in the art, within the scope of the inventive method, can have various variations and change, any variation and the change done, all within protection range of the present invention.
Claims (5)
1. a registration that is applicable to low capacity information carrier is according to method for compressing image, comprise information gathering, image preliminary treatment, frequency domain conversion, quantification and entropy coding, compressed bit stream, it is characterized in that: registration is carried out to specific aim preliminary treatment according to image, image is divided into face, clothing, the several parts of background, different piece is adopted to different processing methods; When coding, the resolution to image, image type, coding parameter are made an appointment, and do not contain any description information of image in image compression code stream, only need a small amount of code stream control information.
2. a kind of registration that is applicable to low capacity information carrier as claimed in claim 1, according to method for compressing image, is characterized in that: concrete treatment step is as follows:
A. in video preview window, predetermined registration, according to the profile of image, is taken pictures by design profile;
B. registration is shone into the preliminary treatment of row image, according to the feature of image in picture, by adopting edge detection algorithm or judge rgb value or the method for gray value, picture is shone to background according to personage's face, registration and clothing is divided into 2-3 part;
C. registration is carried out to dct transform according to image, the human face region in image is carried out to small data piece and cut apart, and fixed block number; Long data block is carried out in all the other regions to be cut apart; For the image of prior agreement resolution, the detection of blocks of data mistake while for the benefit of decoding, the number of blocks of human face region should be fixed, and therefore the image block quantity in other region is also fixed;
D. the data after frequency domain conversion are quantized and entropy coding;
E. in image code stream, data are divided into human face region image block data and other area image blocks of data, human face region image block data is placed on the foremost of code stream, and human face region end-of-data mark is in the end set, utilize end of block flag and human face region end mark, carry out data block automatic error detection.
3. a kind of registration that is applicable to low capacity information carrier as claimed in claim 1 is according to method for compressing image, it is characterized in that: while registration image being divided into background, face, clothing three part according to the dividing method described in step B, background is set as to single value of color or gray value, most of region of clothing is made as to single color-values or gray value.
4. a kind of registration that is applicable to low capacity information carrier as claimed in claim 1 is according to method for compressing image, it is characterized in that: while registration image being divided into background, personage's two parts according to the dividing method described in step B, background is set as to single value of color or gray value.
5. a kind of registration that is applicable to low capacity information carrier as claimed in claim 1 is according to method for compressing image, it is characterized in that: before according to the compressed bit stream arrangement mode described in step D, the compressed bit stream of human face region being come, after the compressed bit stream in other regions comes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410242515.2A CN103986930A (en) | 2014-06-03 | 2014-06-03 | Registration picture image compression method suitable for small-capacity information carrier |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410242515.2A CN103986930A (en) | 2014-06-03 | 2014-06-03 | Registration picture image compression method suitable for small-capacity information carrier |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN103986930A true CN103986930A (en) | 2014-08-13 |
Family
ID=51278742
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410242515.2A Pending CN103986930A (en) | 2014-06-03 | 2014-06-03 | Registration picture image compression method suitable for small-capacity information carrier |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103986930A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109445262A (en) * | 2018-12-28 | 2019-03-08 | 天津利信科技有限公司 | The storage method and device of holographic data |
| WO2019192490A1 (en) * | 2018-04-02 | 2019-10-10 | Huawei Technologies Co., Ltd. | Adaptive quantization in video coding |
| CN113676728A (en) * | 2021-08-25 | 2021-11-19 | 国网吉林省电力有限公司长春供电公司 | Digital platform area unstructured data compression method |
| CN116957524A (en) * | 2023-09-21 | 2023-10-27 | 青岛阿斯顿工程技术转移有限公司 | Talent information intelligent management method and system in technology transfer process |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1779717A (en) * | 2004-11-19 | 2006-05-31 | 公安部第一研究所 | Compression and decompression for certificate photo |
| CN101098473A (en) * | 2006-06-30 | 2008-01-02 | 联想(北京)有限公司 | Picture coding method and apparatus |
-
2014
- 2014-06-03 CN CN201410242515.2A patent/CN103986930A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1779717A (en) * | 2004-11-19 | 2006-05-31 | 公安部第一研究所 | Compression and decompression for certificate photo |
| CN101098473A (en) * | 2006-06-30 | 2008-01-02 | 联想(北京)有限公司 | Picture coding method and apparatus |
Non-Patent Citations (4)
| Title |
|---|
| JIAN-HONG HU ET AL.: "《Compression of personal identification pictures using vector quantization with facial feature correction》", 《OPTICAL ENGINEERING》 * |
| JINCHENG HUANG ET AL.: "《Compression of color facial images using feature correction two-stage vector quantization》", 《IEEE TRANSACTIONS ON IMAGE PROCESSING》 * |
| 张建军: "《身份证件照片分级压缩研究》", 《上海大学》 * |
| 朱秋煜: "《证件照片的特征提取及检索和压缩研究》", 《上海大学》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019192490A1 (en) * | 2018-04-02 | 2019-10-10 | Huawei Technologies Co., Ltd. | Adaptive quantization in video coding |
| CN109445262A (en) * | 2018-12-28 | 2019-03-08 | 天津利信科技有限公司 | The storage method and device of holographic data |
| CN113676728A (en) * | 2021-08-25 | 2021-11-19 | 国网吉林省电力有限公司长春供电公司 | Digital platform area unstructured data compression method |
| CN116957524A (en) * | 2023-09-21 | 2023-10-27 | 青岛阿斯顿工程技术转移有限公司 | Talent information intelligent management method and system in technology transfer process |
| CN116957524B (en) * | 2023-09-21 | 2024-01-05 | 青岛阿斯顿工程技术转移有限公司 | Talent information intelligent management method and system in technology transfer process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101491102B (en) | Video coding considering postprocessing to be performed in the decoder | |
| US6735254B2 (en) | DCT compression using Golomb-Rice coding | |
| CN1186942C (en) | Variance based adaptive block size DCT image compression | |
| CN100576195C (en) | With the system and method for lossless manner to digital picture and voice data decoding | |
| US9258568B2 (en) | Quantization method and apparatus in encoding/decoding | |
| EP0833519B1 (en) | Segmentation and background suppression in JPEG-compressed images using encoding cost data | |
| US20220385944A1 (en) | Non-transform coding | |
| CN103716625A (en) | Image processing apparatus and method | |
| CN112868228B (en) | Self-adaptive DCT sharpener | |
| CN1572119B (en) | Selective chrominance decimation for digital images | |
| CN101653004A (en) | Decoder for selectively decoding predetermined data units from a coded bit stream | |
| CN103986930A (en) | Registration picture image compression method suitable for small-capacity information carrier | |
| CN1937748A (en) | Image compression method | |
| US9106925B2 (en) | WEAV video compression system | |
| CN1539239A (en) | Interface encoding method and apparatus | |
| CN101267566A (en) | Image coding method and device | |
| CN108235038B (en) | Method and device for image coding, decoding and compressing | |
| Singh et al. | JPEG2000: A review and its performance comparison with JPEG | |
| CN101267557B (en) | A method for image compression based on compound vector quantification | |
| CN113422965A (en) | Image compression method and device based on generation countermeasure network | |
| US20070036445A1 (en) | Method and apparatus of efficient lossless data stream coding | |
| Couto | Inpainting-based Image Coding: A Patch-driven Approach | |
| CN100472567C (en) | A method and system of image processing | |
| CN100490534C (en) | Continuous image compression method | |
| Guo et al. | An Improved Image Compression Algorithm for Noise Image |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20140813 |
|
| WD01 | Invention patent application deemed withdrawn after publication |