CN105657385A - Compression method for original image data - Google Patents
Compression method for original image data Download PDFInfo
- Publication number
- CN105657385A CN105657385A CN201511029811.5A CN201511029811A CN105657385A CN 105657385 A CN105657385 A CN 105657385A CN 201511029811 A CN201511029811 A CN 201511029811A CN 105657385 A CN105657385 A CN 105657385A
- Authority
- CN
- China
- Prior art keywords
- image data
- segmentation mark
- raw image
- segmentation
- compression method
- 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.)
- Granted
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N23/00—Cameras or camera modules comprising electronic image sensors; Control thereof
- H04N23/10—Cameras or camera modules comprising electronic image sensors; Control thereof for generating image signals from different wavelengths
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Color Television Systems (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention provides a compression method for original image data. The method comprises the following steps: dividing and reshaping the original image data according to different colors to obtain four color areas; dividing each color area to obtain N sub-areas composed of rectangles or L shapes; converting numerical values in the sub-areas into expression modes of weighted values and a plurality of differences; using the shapes, difference bit widths, weighted values and all differences of the sub-areas to replace the numerical values in the original sub-areas; and carrying out reverse operation to reduce an original image. The compression method provided by the invention is simple in algorithm and is very low requirements on CPU performance, thus being more suitable for hand-held shooting devices mainly relying on embedded system; the storage speed and the image volume can be increased, and no data loss is generated; the sub-matrixes of a various forms can be used for better dividing similar areas; and the compression method provided by the invention is used for effectively compressing the original image data to not only improve the compression rate, but also save the storage time and space and reduce the cost.
Description
Technical field
The present invention relates to imageing sensor and numeral technical field of image processing, be specifically related to a kind of compression method for raw image data.
Background technology
Field of image sensors now, the resolution of imageing sensor is maked rapid progress, the resolution of image and sampling precision growth rate are quickly, sensor more than current 6400W pixel has begun to into people field, and only use the images such as jpeg, pnp that lossy compression method obtains obviously inadequate on this large scale sensor, for professionals, original raw view data has become and indispensable part. And raw image data is very big, with 6400W pixel, 12bit sampling precision is example, one image needs to take at least 96M byte, and imaging time and amount of storage are had significant effect by this by hand-held capture apparatus undoubtedly that commonly use the low storage medium of low capacity (CF card, SD card).
Typically require employing compress technique by Image Data Compression, but, although bzip, winrar of main flow are lossless compress, but the compression ratio of raw image data is non-normally low; Although JPG, PNP etc. are high for the compression algorithm compression ratio of image, but are all lossy compression method, it is impossible to recover initial data; Although JPEG2000 is lossless compress, but algorithm is complicated, the image operation meeting up to a hundred per second surely even high performance CPU also differs, and JPEG2000 does not support original bayerraw view data.
Summary of the invention
In order to overcome problem above, the invention provides a kind of compression algorithm for raw image data, improve data compression without loss rate and time.
In order to achieve the above object, the invention provides the compression method for raw image data, including:
Step 01: raw image data is carried out segmentation shaping according to different colors, obtains four color regions;
Step 02: each color region being split and obtains N number of subregion being made up of rectangle or L-shaped, N is be more than or equal to 1;
Step 03: the numerical value in described subregion is converted into the representation adopting weighted value and multiple difference;
Step 04: use the shape of subregion, difference bit wide, weighted value and all of difference to replace numerical value original in described subregion;
Step 05: carry out reverse operation, restore original image.
Preferably, the storage format of described raw image data is bayer, and described step 01 specifically includes: tetra-kinds of color extraction of the RGGB in described raw image data are out formed a R color region, two G color regions and a B color region.
Preferably, each described color region is square M by M matrix.
Preferably, first square matrix is divided into the matrix of X KXK, described K is 4, matrix to each 4 �� 4 continues segmentation formation 1,2 or 4 submatrixs, the shape one of submatrix has eight, corresponding eight different segmentations mark, respectively: the first segmentation mark (000), the second segmentation mark (100), the 3rd segmentation mark (101), the 4th segmentation mark (110), the 5th segmentation mark (111), the 6th segmentation mark (010), the 7th segmentation mark (011) and the 8th segmentation mark (001).
Preferably, described step 02 meets following rule when being split by described square matrix: when the row and column of square matrix is unsatisfactory for the multiple of 4, by the row and column polishing of square matrix to the multiple meeting 4; 8th segmentation mark (001) can only 4 composition 4 �� 4 matrix occurs together; 7th segmentation mark (011) must occur composition 4 �� 4 matrix in pairs, and order is first up and then down; It is converted into a submatrix during peripheral 7 pixels calculating of the second segmentation mark (100), the 3rd segmentation mark (101), the 4th segmentation mark (110) and the 5th segmentation mark (111), the order of pixel is first left and then right, then from top to bottom; The priority of described segmentation mark is: the first segmentation mark (000) > the second segmentation mark (100)=3rd segmentation mark (101)=4th segmentation mark (110)=5th segmentation mark (111) > the 6th segmentation mark (010)=7th segmentation mark (011) > the 8th segmentation mark (001).
Preferably, described step 03 specifically includes: numerical value minimum in each numerical value of each submatrix is used as weighted value, other numerical value and the difference of this weighted value replace original numerical value, add the shape of submatrix, difference bit wide, and the submatrix after being changed.
Preferably, described step 03 also includes: before converting, and selects the shape of described submatrix, according to priority order, selects the shape that compression ratio is the highest.
Preferably, in described step 04, form a complete header file structure according to the AD sampling precision of the length of original image and wide, original color arrangement mode and original image.
Preferably, described step 05 also includes: when the row or column of the square matrix calculated is unsatisfactory for the multiple of K, use last row or column in described header file structure to fill row or column further and become the multiple of K to calculate to obtain original image.
Preferably, the half of the row of raw image data described in the behavior of square matrix, it is classified as the half of the row of described raw image data.
The compression algorithm for raw image data of the present invention, by raw image data shaping, region segmentation, regioinvertions, and the formation of header file structure, thus restoring original image; The algorithm of the present invention is simple, requires extremely low to cpu performance, is particularly suited for the hand-held capture apparatus that embedded system is master;Storage speed and image volume can be improved, and data do not have any loss, it is possible to restore raw image data; It is different from Normal squeezing algorithm and uses the square matrix segmentation of M �� M, use the submatrix of variform can better mark off similar area; Present invention achieves being effectively compressed raw image data, not only increase compression ratio, also a saving storage time and space, reduce cost.
Accompanying drawing explanation
Fig. 1 is the schematic diagram of the bayer picture format of a preferred embodiment of the present invention
Fig. 2 is the schematic diagram of the array formed after the extraction same pixel point of a preferred embodiment of the present invention
Fig. 3 is eight kinds of region segmentation pictures of a preferred embodiment of the present invention
Fig. 4 is the segmentation schematic diagram of the square matrix of a preferred embodiment of the present invention
Fig. 5 is the schematic diagram of the submatrix that the square matrix of a preferred embodiment of the present invention is divided into
Fig. 6 is the schematic diagram of the submatrix after the conversion of a preferred embodiment of the present invention
Fig. 7 is the schematic flow sheet of the compression algorithm for raw image data of a preferred embodiment of the present invention
Detailed description of the invention
For making present disclosure clearly understandable, below in conjunction with Figure of description, present disclosure is described further. Certainly the invention is not limited in this specific embodiment, the general replacement known by those skilled in the art is also covered by protection scope of the present invention.
The feature of view data is exactly regional, the same area color difference is only small, but it is general without identical, on equipment particularly in high AD sampling precision, this algorithm is mainly according to this feature of view data, divide the image into into multiple little region, one group of little for numerical value difference in same region numerical value is simplified to: the mode of weighted number+difference, because in most cases in the same area, the numerical value difference of each pixel is only small, the figure place that the figure place of difference takies much smaller than pixel value itself, reaches the purpose of compression size of data with this.
Below in conjunction with accompanying drawing 1-6 and specific embodiment, the present invention is described in further detail. It should be noted that, accompanying drawing all adopts the form simplified very much, uses non-ratio accurately, and only in order to conveniently, clearly to reach to aid in illustrating the purpose of the present embodiment.
Step 01: raw image data is carried out segmentation shaping according to different colors, obtains four kinds of color regions;
Concrete, the storage format of raw image data is bayer, specifically includes: tetra-kinds of color extraction of the RGGB in described raw image data are out formed a R color region, two G color regions and a B color region, each color region is square M by M matrix, moment sensor commonly uses bayer picture format to store raw image data, because for coloured image, need to gather multiple most basic color, such as tri-kinds of colors of RGB, simplest method is exactly by the method for filter, red filter is through red wavelength, green filter is through green wavelength, blue filter is through blue wavelength, if gathering tri-Essential colour of RGB, then need three pieces of filters, so expensive, and bad manufacture, because three pieces of filters all must assure that each pixel aligns, when with bayer form time, well solve this problem, the different color that bayer format picture is arranged on one piece of filter, by analyzing human eye, the perception of color is found, human eye is more sensitive to green, so the pixel of the picture green form of general bayer form be R and G pixel and, as shown in Figure 1.
Clearly this structure inapplicable because adjacent point represents the different color of RGB respectively, numerical value differs greatly, so the first step we by RGGB these four color extraction out, form 4 Founders as shown in Figure 2; Although it is noted that image is RGGB, but this algorithm can be suitable for the color combination of various bayer form. Understand in order to convenient, by raw image data array called after matrix L, after first step shaping, form 4 square matrix called after J1, J2, J3 and J4.
Step 02: each color region being split and obtains N number of subregion being made up of rectangle or L-shaped, N is be more than or equal to 1;
Concrete, for square matrix J1 (other regions all use same method). First MXM square matrix is divided into the matrix of X K �� K; When K is 4, then the matrix of 4 �� 4 is divided into 1,2 or 4 submatrix, each submatrix called after j; Submatrix j is made up of rectangle or L-shaped; The difference that optimal situation is 16 pixels in submatrix j is all only small, it is possible to directly as calculating matrix, the submatrix j of 4 �� 4 is come computing; Certain practical situation will not be all so desirable, so this algorithm provides altogether the dividing method of 8 kinds of square matrixs, uses the expression position of 3bit to represent this 8 kinds of partitioning schemes, this 3bit called after segmentation mark; The shape one of submatrix j has eight, the segmentation mark of corresponding difform submatrix j is also eight altogether, respectively: the first segmentation mark (000), the second segmentation mark (100), the 3rd segmentation mark (101), the 4th segmentation mark (110), the 5th segmentation mark (111), the 6th segmentation mark (010), the 7th segmentation mark (011) and the 8th segmentation mark (001), as shown in Figure 3; In Fig. 3, order from left to right and from top to bottom segmentation mark constitute submatrix j combination respectively 4 �� 4 first segmentation mark (000), second segmentation mark (100) of 3 �� 3+7,3rd segmentation mark (101) of 3 �� 3+7,4th segmentation mark (110) of 3 �� 3+7,5th segmentation mark (111) of 3 �� 3+7,6th segmentation mark (010) of 4 �� 2,7th segmentation mark (011) of 2 �� 4, the 8th segmentation mark (001) of 2 �� 2.
In order to allow the square matrix J of arbitrary size that different submatrix j can be divided into flexibly to combine, algorithm also has following several agreements:
When the row and column of square matrix J is unsatisfactory for the multiple of 4, by the row and column polishing of square matrix J to the multiple meeting 4;
8th segmentation mark (001) can only 4 composition 4 �� 4 submatrix j occurs together; 7th segmentation mark (011) must occur composition 4 �� 4 submatrix j in pairs, and order is first up and then down;
7 pixels of the second segmentation mark (100), the 3rd segmentation mark (101), the 4th segmentation mark (110) and the 5th segmentation mark (111) peripheral (peripheries of these 4 kinds of 3x3 matrixes) are converted into a submatrix j when calculating, the order of pixel is first left and then right, then from top to bottom;
The priority of segmentation mark is: the first segmentation mark (000) > the second segmentation mark (100)=3rd segmentation mark (101)=4th segmentation mark (110)=5th segmentation mark (111) > the 6th segmentation mark (010)=7th segmentation mark (011) > the 8th segmentation mark (001), namely preferentially use the matrix of maximum area to combine.
So, square matrix J is partitioned into the combination of the submatrix j of various different shapes as shown in Figure 4;
Step 03: the numerical value in subregion is converted into the representation adopting weighted value and multiple difference;
Concrete, before converting, select the shape of submatrix j, according to priority order, select the shape that compression ratio is the highest; Then, 1 or multiple matrix j are converted; Numerical value minimum in each numerical value of j is used as weighted value, and other numerical value and the difference of this weighted value replace original numerical value and obtain difference, add the shape of submatrix, difference bit wide, the submatrix j after being changed;
16 numerical value inside each submatrix j being converted into the representation of a minimum value and multiple difference, first find a minimum numerical value as weighted value, other each numerical value all replace original numerical value by difference. Because the particularity of view data, the pixel value difference of adjacent area is only small, such as the picture of 14 sampling precisions, and the difference of the same area is possibly less than 5, the difference using 5 replaces the initial data of 14 can be substantially reduced figure place, thus reaching the purpose of lossless compress.
In the present embodiment, submatrix j after conversion is divided into three region ABCD1D2D3D4 ... B wherein CD1D2D3.....A is matrix shape, B is difference width, one has 3 bit, represents that from 000 to 111 the figure place of difference is respectively: 3,4,5,6,7,8,9,10, C represents weighted value, bit wide is the same with the bit wide of initial data, D1, D2, D3 etc. represent the difference of each pixel respectively, and bit wide is equal to the value of difference mark instruction, and the order of pixel is from left to right; If comprising 2 or 4 regions in form, then areas combine below is BCD1, D2, D3..... (not needing A shape mark)
Refer to Fig. 5, for the submatrix j of a 4x4, (AD sampling precision is 14) is described;
Converting according to this algorithm: in 16 numerals, minima is 1505, using 1505 as weighted value, the difference calculating each pixel value and weighted value is shown in table one:
Table one
Refer to Fig. 6, maximum difference width is 6, so difference bit width is designated 011b (width of difference is 6bit), weighted value is 00010111100001b (1505), the difference of first pixel is 011100b (28), second pixel is 000000b (0) ... ... ... the 16th pixel is 110110b (54), sum total bit=3+3+14+6 �� 16=116 after conversion, initial data total bit is 14 �� 16=224, and compression ratio is 116/224=51%.
It is also possible to calculate the compression ratio of other a few Seed Matrix j under 14bit data bit width, as shown in Table 2:
Table two
3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | |
4x4 | 29 | 36 | 43 | 51 | 57 | 64 | 71 | 78 |
3x3 | 37 | 44 | 51 | 58 | 65 | 73 | 80 | 87 |
2x4,4x2 | 39 | 46 | 53 | 60 | 67 | 75 | 82 | 89 |
3x3 is peripheral | 41 | 48 | 56 | 63 | 70 | 77 | 84 | |
2x2 | 57 | 64 | 71 | 78 | 85 | 92 |
The compression ratio of the submatrix j calculated when AD sampling precision is 12, as shown in Table 3;
Table three
The compression ratio of the submatrix j calculated when AD sampling precision is 16, as shown in Table 4:
3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | |
4x4 | 30 | 37 | 44 | 51 | 58 | 66 | 73 | 80 |
3x3 | 37 | 44 | 51 | 58 | 65 | 73 | 80 | 87 |
2x4,4x2 | 39 | 46 | 53 | 60 | 67 | 75 | 82 | 89 |
3x3 is peripheral | 41 | 48 | 56 | 63 | 70 | 77 | 84 | 91 |
2x2 | 57 | 64 | 71 | 78 | 85 | 92 |
From upper table it will be seen that the AD sampling precision of image is more high, compression ratio is more big, same sampling precision, and the area of submatrix j is more big, and compression ratio is more big, so when the shape of submatrix j, according to priority order, selecting the submatrix j that compression ratio is the highest.
In practical application, although AD sampling precision is more big, compression ratio is about big, but the fluctuation range of numerical value is also relatively large, equally, the submatrix j that area is bigger, because the pixel comprised is more, so the bit wide of difference is also relatively large, the submatrix j that area is little, such as contain the submatrix j of 2 �� 2 the 8th segmentations mark (001), because only that 4 points, be readily available less difference figure place on the contrary.
So generally, the compression ratio of RAW image data is approximately 40% to 60% (view sheet complexity and change) by this algorithm.
Step 04: use the shape of subregion, difference bit wide, weighted value and all of difference to replace numerical value original in subregion;
Concrete, form a complete header file structure according to the AD sampling precision of the length of original image and wide, original color arrangement mode and original image, as shown in Table 5:
Here, row and column in header file structure is the row and column of raw image data, the namely row and column size of aforesaid raw image data array A, the half of the row of the square matrix J row equal to raw image data array A, the half of the row of the square matrix J row equal to raw image data array A.
Step 05: carry out reverse operation, restore original image.
Concrete, when the row or column of the K �� K square matrix J calculated is unsatisfactory for the multiple of K, uses last row or column in header file structure to fill row or column further and become the multiple of K to calculate to obtain original image. Here, K is 4, namely when the row and column of the square matrix calculated is unsatisfactory for the multiple of 4, it is necessary to uses last row and column filling head file structure to make the row and column of square matrix meet the multiple of 4, then carries out computing again.
Such as, the row of the header file struc-ture of one file is equal to 3404, row are equal to 2804, then the actual row and column of square matrix B is equal to 1702 �� 1402, but the square matrix J after converting is sized to 1704 �� 1404, so when this file is reduced into bayer original image, it is first according to the size of 1704 �� 1404 and reduces 4 square matrix J, last 2 row and 2 row are abandoned after reduction, obtain the square matrix B of 1702 �� 1404, finally 4 square matrix J are changed position and obtain the bayer original image of 3404 �� 2804.
Although the present invention discloses as above with preferred embodiment; right described embodiment is illustrated only for the purposes of explanation; it is not limited to the present invention; those skilled in the art can do some changes and retouching without departing from the spirit and scope of the present invention, and the protection domain that the present invention advocates should be as the criterion with described in claims.
Claims (10)
1. the compression method for raw image data, it is characterised in that including:
Step 01: raw image data is carried out segmentation shaping according to different colors, obtains four color regions;
Step 02: each color region being split and obtains N number of subregion being made up of rectangle or L-shaped, N is be more than or equal to 1;
Step 03: the numerical value in described subregion is converted into the representation adopting weighted value and multiple difference;
Step 04: use the shape of subregion, difference bit wide, weighted value and all of difference to replace numerical value original in described subregion;
Step 05: carry out reverse operation, restore original image.
2. the compression method for raw image data according to claim 1, it is characterized in that, the storage format of described raw image data is bayer, and described step 01 specifically includes: tetra-kinds of color extraction of the RGGB in described raw image data are out formed a R color region, two G color regions and a B color region.
3. the compression method for raw image data according to claim 1, it is characterised in that each described color region is square M by M matrix.
4. the compression method for raw image data according to claim 3, it is characterized in that, first square M by M Factorization algorithm is become the matrix of X K �� K, described K is 4, matrix to each 4 �� 4 continues segmentation formation one, two or four submatrixs, the shape one of submatrix has eight, corresponding eight different segmentations mark, respectively: the first segmentation mark (000), second segmentation mark (100), 3rd segmentation mark (101), 4th segmentation mark (110), 5th segmentation mark (111), 6th segmentation mark (010), 7th segmentation mark (011) and the 8th segmentation mark (001).
5. the compression method for raw image data according to claim 4, it is characterized in that, described step 02 meets following rule when being split by described square matrix: when the row and column of square matrix is unsatisfactory for the multiple of 4, by the row and column polishing of square matrix to the multiple meeting 4; 8th segmentation mark (001) can only 4 composition 4 �� 4 matrix occurs together; 7th segmentation mark (011) must occur composition 4 �� 4 matrix in pairs, and order is first up and then down; It is converted into a submatrix during peripheral 7 pixels calculating of the second segmentation mark (100), the 3rd segmentation mark (101), the 4th segmentation mark (110) and the 5th segmentation mark (111), the order of pixel is first left and then right, then from top to bottom; The priority of described segmentation mark is: the first segmentation mark (000) > the second segmentation mark (100)=3rd segmentation mark (101)=4th segmentation mark (110)=5th segmentation mark (111) > the 6th segmentation mark (010)=7th segmentation mark (011) > the 8th segmentation mark (001).
6. the compression method for raw image data according to claim 3, it is characterized in that, described step 03 specifically includes: numerical value minimum in each numerical value of each submatrix is used as weighted value, other numerical value and the difference of this weighted value replace original numerical value, add the shape of submatrix, difference bit wide, and the submatrix after being changed.
7. the compression method for raw image data according to claim 6, it is characterised in that described step 03 also includes: before converting, and selects the shape of described submatrix, according to priority order, selects the shape that compression ratio is the highest.
8. the compression method for raw image data according to claim 1, it is characterized in that, in described step 04, form a complete header file structure according to the AD sampling precision of the length of original image and wide, original color arrangement mode and original image.
9. the compression method for raw image data according to claim 8, it is characterized in that, described step 05 also includes: when the row or column of the square M by M matrix calculated is unsatisfactory for the multiple of K, use last row or column in described header file structure to fill row or column further and become the multiple of K to calculate to obtain original image.
10. the compression method for raw image data according to claim 3, it is characterised in that the half of the row of raw image data described in the behavior of square M by M matrix, is classified as the half of the row of described raw image data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201511029811.5A CN105657385B (en) | 2015-12-31 | 2015-12-31 | A kind of compression method for raw image data |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201511029811.5A CN105657385B (en) | 2015-12-31 | 2015-12-31 | A kind of compression method for raw image data |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105657385A true CN105657385A (en) | 2016-06-08 |
CN105657385B CN105657385B (en) | 2018-02-23 |
Family
ID=56491144
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201511029811.5A Active CN105657385B (en) | 2015-12-31 | 2015-12-31 | A kind of compression method for raw image data |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105657385B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111711511A (en) * | 2020-06-16 | 2020-09-25 | 电子科技大学 | Method for lossy compression of frequency domain data |
CN114567782A (en) * | 2022-04-27 | 2022-05-31 | 江苏游隼微电子有限公司 | Raw image compression method and device suitable for 3DNR image noise reduction |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1630371A (en) * | 2003-12-19 | 2005-06-22 | 联想(北京)有限公司 | A method of lossless image compression applied to real time transmission |
CN1674665A (en) * | 2004-03-26 | 2005-09-28 | 奥林巴斯株式会社 | Image compressing method and image compression apparatus |
CN1716998A (en) * | 2004-06-14 | 2006-01-04 | 奥林巴斯株式会社 | Picture compressor and picture reproducer |
CN1764232A (en) * | 2004-10-19 | 2006-04-26 | 微软公司 | System and method for encoding mosaiced image data employing a reversible color transform |
US7469059B1 (en) * | 2004-12-17 | 2008-12-23 | Google Inc. | Reorganization of raw image data for processing |
CN101753780A (en) * | 2009-12-31 | 2010-06-23 | 中国科学院西安光学精密机械研究所 | Compression method of color image based on primary color extraction, segmentation and compression |
CN103414903A (en) * | 2013-08-28 | 2013-11-27 | 清华大学 | Compressing method and device for Bayer format images |
-
2015
- 2015-12-31 CN CN201511029811.5A patent/CN105657385B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1630371A (en) * | 2003-12-19 | 2005-06-22 | 联想(北京)有限公司 | A method of lossless image compression applied to real time transmission |
CN1674665A (en) * | 2004-03-26 | 2005-09-28 | 奥林巴斯株式会社 | Image compressing method and image compression apparatus |
CN1716998A (en) * | 2004-06-14 | 2006-01-04 | 奥林巴斯株式会社 | Picture compressor and picture reproducer |
CN1764232A (en) * | 2004-10-19 | 2006-04-26 | 微软公司 | System and method for encoding mosaiced image data employing a reversible color transform |
US7469059B1 (en) * | 2004-12-17 | 2008-12-23 | Google Inc. | Reorganization of raw image data for processing |
CN101753780A (en) * | 2009-12-31 | 2010-06-23 | 中国科学院西安光学精密机械研究所 | Compression method of color image based on primary color extraction, segmentation and compression |
CN103414903A (en) * | 2013-08-28 | 2013-11-27 | 清华大学 | Compressing method and device for Bayer format images |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111711511A (en) * | 2020-06-16 | 2020-09-25 | 电子科技大学 | Method for lossy compression of frequency domain data |
CN111711511B (en) * | 2020-06-16 | 2021-07-13 | 电子科技大学 | Method for lossy compression of frequency domain data |
CN114567782A (en) * | 2022-04-27 | 2022-05-31 | 江苏游隼微电子有限公司 | Raw image compression method and device suitable for 3DNR image noise reduction |
CN114567782B (en) * | 2022-04-27 | 2022-07-12 | 江苏游隼微电子有限公司 | Raw image compression method and device suitable for 3DNR image noise reduction |
Also Published As
Publication number | Publication date |
---|---|
CN105657385B (en) | 2018-02-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2015120823A1 (en) | Image compression method and device using reference pixel storage space in multiple forms | |
CN1235414C (en) | Apparatus and method for compressing the picture data received from picture sensor | |
CN108141505B (en) | Compression and decompression method for high bit depth medical gray level image | |
CN102881026B (en) | Palette device and generation method for image with transparency information | |
WO2016172994A1 (en) | Image coding and decoding method and device | |
US20130251261A1 (en) | Method And Apparatus For Image Data Compression | |
US11653019B2 (en) | Image coding and decoding methods, image processing device and computer storage medium | |
WO2002067589A1 (en) | Image processing system, image processing method, and image processing program | |
EP3792908A1 (en) | Data compression system and data mapping method | |
CN112422982A (en) | Random accessible image data compression | |
JP6794364B2 (en) | Image coding method and device and image decoding method and device | |
CN111405294B (en) | Storage and transmission method of image raster data | |
CN105657385A (en) | Compression method for original image data | |
CN108632610A (en) | A kind of colour image compression method based on interpolation reconstruction | |
US8712170B2 (en) | Image-processing method and program, and image-processing apparatus | |
JP4780215B2 (en) | Image processing apparatus and image processing program | |
CN102118616A (en) | Picture decoding method and picture decoder | |
CN108235024B (en) | Method and device for compressing image | |
US20160284056A1 (en) | Image processing apparatus and method | |
WO2020042467A1 (en) | Image compression method, apparatus and device, and computer-readable storage medium | |
JP5933987B2 (en) | Image coding apparatus and control method thereof | |
JP2015061214A (en) | Image compression device, image compression method, program and recording medium | |
CN108200433B (en) | Image compression and decompression method | |
CN113452995A (en) | Data coding and decoding method and device with different scanning directions of current string and reference string | |
CN106296754B (en) | Show data compression method and display data processing system |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |