JP4520197B2 - Image data compression method - Google Patents

Description

  The present invention relates to a method for compressing image data.

As an image compression method in which the quantization rate to a user's favorite place is variable at the time of image compression, the method described in Patent Document 1 is known. In this conventional example, the image is decomposed into RGB or YUV components. Thereafter, each component is divided into tile regions, and then wavelet transformed with respect to each of the components. Further, each component subjected to wavelet transformation is encoded with a different quantization rate.
JP 2003-339047 A

  However, since the above-described conventional technique adjusts the dot density of the restored image by adjusting the quantization rate, it assumes lossy compression and has a problem of lack of restoration accuracy.

  The present invention has been made to solve the above-described drawbacks, and an object of the present invention is to provide a method for compressing image data that is reversible and has high compression efficiency.

According to the present invention, the object is
A method of compressing image data in which a digital image is grouped and compressed by a pixel value difference with an adjacent pixel ,
Dividing the image into a predetermined number of color planes;
Each color plane
Only a pixel whose pixel value difference with an adjacent pixel exceeds a predetermined threshold with respect to a certain search direction is extracted as a feature pixel 1, and a feature pixel block 2 consisting of a pixel group of only the feature pixel 1 ;
The remaining pixels associated with each feature pixel 1 are extracted as auxiliary pixels 3 and grouped into auxiliary pixel blocks 4 consisting of only the auxiliary pixels 3 ,
This is achieved by providing a method for compressing image data in which pixel information is converted and encoded in units of blocks.

  In image compression, first, a digital image is divided into a predetermined number of color planes. The number of color planes varies depending on the nature of the image, one for a monochrome two-tone or grayscale image, three for a color image in R, G, B mode, Y, Cb, Cr mode, C, M, Y, The number of K-mode color images is four.

  The compression is performed by grouping each color plane into the characteristic pixel block 2 and the auxiliary pixel block 4. The feature pixel 1 is an end point of continuity in a pixel row or a planar pixel group, and is irreversible in the sense that an image can be reproduced by appropriately interpolating between feature points at the time of restoration. It can also be used as a source for compression.

  On the other hand, the auxiliary pixel 3 is a pixel group inserted between the feature pixels 1 and 1 and can have various positions corresponding to the position with respect to the feature pixel 1. That is, when irreversible compression is assumed, compression for the auxiliary pixel 3 group can be made irreversible, or the auxiliary pixel 3 can be omitted to improve the compression efficiency. On the other hand, when lossless compression is required, that is, when accurate image reproduction is required, lossless compression may be performed on the auxiliary pixel block 4 as well.

  As described above, in the present invention in which an image is grouped into the characteristic pixel block 2 and the auxiliary pixel block 4 and the pixel value is independently converted for each of the feature pixel block 2 and the auxiliary pixel block 4, various steps are performed from the lossless compression depending on the type of the image. It is possible to arbitrarily select irreversible compression, and versatility is dramatically improved.

  In the present invention, the feature pixel block 2 is common in that isolation of the pixel value from the subsequent auxiliary pixel 3 is seen, and the difference between the discrete points is not necessarily large. 4 is selected on the basis of the closeness of pixel values, so the difference between adjacent pixels is small. Therefore, when pixel values are converted using this feature, the number of bits required for pixel value description can be reduced, and compression efficiency is improved.

In this case, if the auxiliary pixel 3 continues for a predetermined length while changing smoothly, the difference between a pair of characteristic pixel 1 values sandwiching the auxiliary pixel 3 group may be excessive, and the auxiliary pixel block 4 description data length becomes excessive, because in some cases the format efficiency is deteriorated, when the auxiliary pixel 3 consecutive predetermined length to the search direction, the last of the auxiliary pixel 3 forms a predetermined length continuously in the search direction Is preferably extracted as the feature pixel 1.

Pixel information for the feature pixel 1, that is, information on the position and the pixel value is
It is possible to reduce the number of description bits by describing with the converted value to the relative position information from the other immediately preceding characteristic pixel 1. The relative position information can be expressed by a relative position in the image space consisting of a pixel matrix-pixel value (gradation), that is, (pixel distance, pixel value difference) or (pixel distance, pixel value angle). .

Furthermore, for the auxiliary pixel 3,
The description bit number can be reduced by describing the corresponding feature pixel 1 or the converted value into the difference from the pixel value of the other auxiliary pixel 3 just before.

  In addition, the pixel value of the feature pixel 1 belongs to another pixel column parallel to the search direction, and the distance between the two pixels matches the pixel value of another feature pixel 1 that is equal to or less than a predetermined threshold value. The pixel value of the feature pixel 1 can be described by a conversion value into flag information indicating the presence of the feature pixel 1 referred to in another column.

The pixel search is performed along the scanning direction or the sub-scanning direction at the time of image acquisition. When the end of the pixel column is reached, the pixel search is performed on a pixel column adjacent to such a search pixel column.
The compression method of image data is
After dividing the image into a plurality of image blocks,
Each divided image block 5 is compressed by the method according to claim 1 or 2,
Thereafter, a compression method of image data to be combined and encoded by adding image block content information to the head of each divided image block 5,
The image block content information includes the compression mode of the auxiliary pixel block in addition to the number of each divided image block, the position in the image, the size information,
The compression mode can be configured to be set to either compression in the lossless mode of the auxiliary pixel block, non-use, or compression in the lossy mode .

  By changing the degree of compression for each divided image block 5, it is possible to freely select and compress an area in the image that emphasizes fidelity to the original image and an area that emphasizes restoration speed in consideration of the restoration environment. Can be compressed.

  According to the present invention, the compression efficiency can be increased even if it is reversible.

  FIG. 1A shows a configuration example of the image compression apparatus. This image compression apparatus can be realized by running a program that causes a computer to function as each unit described later on the computer.

  In this block diagram, 6 is an input unit and 7 is a block dividing means. The block dividing means 7 divides the input digital image into a plurality of rectangular blocks, and generates a divided image block 5 (step Sc1) (see FIG. 2A). For the block dividing means 7, the number of divisions, the compression mode for each divided image block 5, etc. can be set. Further, the arrangement of the divided image blocks 5 does not necessarily follow a certain order, for example, as arranged in a certain direction from the upper left, and can be arranged randomly to improve security. Factors for generating these divided blocks are added to each divided image block 5 as a block header 8 and output to the subsequent color plane dividing means 9 as described later.

The color plane dividing means 9 divides the image into a suitable number of color planes (color channels) according to the image mode of each divided image block 5 (step Sc2).
For each color plane divided as described above, the feature pixel 1 is first extracted by the feature pixel extraction means 10 (step Sc3). FIG. 3 shows an example of a flowchart for extracting the feature pixel 1 by searching the color plane of p rows and q columns in the row direction. In FIG. 3, the suffix “L” is the number of rows in the image, “C” is the number of columns, “n” is the serial number of the feature pixel 1, and the feature pixel v (n) is assumed for simplicity of explanation. An array for storing the pixel information of the feature pixel 1 and the auxiliary pixel v (n, m) indicate an array for storing the pixel information of the mth auxiliary pixel 3 to which the nth feature pixel 1 is subordinate. Further, in this example, when extracting the feature pixel 1, the pixel value conversion of the feature pixel 1 and the auxiliary pixel 3 is simultaneously performed on the premise of lossless compression. However, the auxiliary pixel 3 is irreversibly compressed. In this case, the stored value in the auxiliary pixel v is appropriately changed in consideration of the efficiency of pixel information conversion at the time of lossy compression.

  When extracting the feature pixel 1, first, the pixel in the first row and the first column is extracted as the feature pixel 1, the pixel value is stored in the feature pixel v (n) (step Sc3-1), and the search target is the pixel in the next column. (Step Sc3-2). If the next pixel is present in the search row (step Sc3-3), the difference between the pixel value of the previous pixel and the pixel value of the search target pixel is calculated and compared with a predetermined threshold (Th) ( Step Sc3-4) When the difference (ΔV) is equal to or smaller than the threshold value, the auxiliary pixel 3 is extracted, and when it is equal to or larger than the threshold value, the characteristic pixel 1 is extracted.

  FIG. 4A shows the extraction state of the feature pixel 1 and the auxiliary pixel 3. In FIG. 4A, the vertical axis indicates the pixel value V (gradation), the horizontal axis indicates the pixel column position, and when the pixel in the leftmost column is the feature pixel 1, the difference from the adjacent pixel (ΔV1)> When Th is assumed, the adjacent pixel is extracted as the feature pixel 1. 4, 5, and 6, the feature pixel 1 is indicated by a black circle and the auxiliary pixel 3 is indicated by a white circle. On the other hand, if the difference (ΔV2) between the pixel in the second column (feature pixel 1) is smaller than Th, the pixel in the third column is extracted as the auxiliary pixel 3, and in the same manner, The pixel values are sorted by comparing the pixel value difference (ΔV) between the two and Th.

  In this embodiment, as shown in FIG. 4A, when the auxiliary pixel 3 continues excessively, the auxiliary pixel 3 is registered as the feature pixel 1. Therefore, when storing in the auxiliary pixel v, first, it is confirmed that the number of consecutive auxiliary pixels 3 is equal to or less than a predetermined number (j) (step Sc3-5), and only when the condition is satisfied. The difference value (ΔV = ΔP) is stored in the auxiliary pixel v (n, m) (step Sc3-6).

  In this embodiment, the stored value in the auxiliary pixel v, that is, the conversion value of the pixel information is set to the difference (ΔP) with the adjacent pixel as shown in FIG. As shown in FIG. 5B, the difference (ΔP) from the related feature pixel 1 may be used.

  The exclusion operation of the auxiliary pixel 3 for extracting the characteristic pixel 1 is performed by checking whether the number of consecutive auxiliary pixels 3 reaches j, whether the pixel value difference with an adjacent pixel exceeds a predetermined threshold (Th), Until one of the search ends, it is performed. In this embodiment, the row start column, that is, (L, 1) is set to be the feature pixel 1 unconditionally, but it is determined by the difference from the immediately preceding survey pixel (L-1, q). You can also. In this case, as illustrated, the search direction does not move from (L, q) to (L + 1, 1), but instead moves to (L + 1, q), and thereafter (L + 1, q-1), ( L + 1, q-2)... (L + 1, 1) may be searched in the reverse direction, and a zigzag search as a whole may be performed.

  When the feature pixel 1 is detected, first, the possibility of reference to another row is detected (step Sc3-7). Reference to another row refers to the case where the pixel to be examined is the feature pixel 1, searches for pixels in the same column retroactively by a predetermined number of rows (i rows), and includes the feature pixel 1 having the same pixel value among them. In this case, the pixel value of the pixel to be investigated is described by referring to another feature pixel 1.

  Reference to another row will be described with reference to FIG. Now, in FIG. 4B, when the feature pixel 1 is detected in the first column of the nth row (Ln), if the i value is 1, then the pixel (Ln− It is determined whether or not 1,1) is the feature pixel 1. In this example, since the pixel (Ln-1, 1) is displayed as a black circle, it is the feature pixel 1, and when this condition is satisfied, it is next determined whether the pixel values match. In this example, since the pixel value (Ln-1, 1) <> pixel value (Ln, 1) does not satisfy the condition, the pixel information for the pixel (Ln, 1) is referred to another row. Never used. Similarly, in FIG. 4B, since the pixel (Ln, 2) is the feature pixel 1, but the pixel (Ln-1, 2) is not the feature pixel 1, no reference to another row is used. On the other hand, since both the pixel (Ln, Cm) and (Ln-1, Cm) are the feature pixel 1 and the pixel values are the same, the pixel information for the pixel (Ln, Cm) Is described using a reference to another line, and in this embodiment, it is indicated that a reference to another line is performed by describing “0” (step Sc3-8).

  On the other hand, if there is no feature pixel 1 to be referenced, the relative position information for the feature pixel extracted immediately before is stored in the feature pixel v (step Sc3-9). The pixel information for the feature pixel 1 may be indicated by the number of pixels (distance ΔX) from the immediately preceding feature pixel 1 and the pixel value difference (ΔD), as shown in FIG. ) As shown by ΔX and an angle (θ).

  When the extraction of the feature pixel 1 is completed for all the pixels as described above, a format is generated in the format generation means 11 (step Sc4). As shown in FIG. 7, the format generation is a format conversion step of grouping each of the divided image blocks 5 into the characteristic pixel block 2 and the auxiliary pixel block 4 prior to encoding. Are divided into a sub-block in which the number of pixels (ΔX) in the search direction between adjacent feature pixels 1 and 1 is collected, and a sub-block consisting of a pixel value difference (ΔD) between adjacent feature pixels 1 and 1. Regarding the auxiliary pixel 3, since the interval between the adjacent auxiliary pixels 3 and 3 is determined to be '1' dots, it is described only by the pixel value difference (ΔP) between the adjacent auxiliary pixels 3.

As described above, when the lossy compression is selected for the divided image block 5 to be processed, the information regarding the auxiliary pixel 3 is described with contents other than (ΔP).
As described above, grouping by sub-blocks is performed on each color plane, and then synthesized and encoded by the encoding means 12 (step Sc5). For encoding, Huffman coding or entropy coding such as arithmetic coding is used. As shown in FIG. 7, a sub-block header 13 and a sub-block end mark 14 are added to the beginning and end of each sub-block compressed in this way, and the encoding step ends.

  When the encoding step is completed, the image block synthesizing unit 15 executes the synthesis step (Sc6) of the divided image block 5. As shown in FIG. 7, the composition of the divided image block 5 is a process of giving a block header 8 and a block end mark 16 between the compressed divided image blocks 5 and outputting them from the output unit 17 as one file. In the block header 8, the compression mode for the auxiliary pixel 3 is described in addition to the number of the divided image block 5, the position in the image, and the size information.

  FIG. 1 (b) shows the decompression apparatus for the compressed file generated as described above, and FIG. 2 (b) shows the operation flow. Similar to the image compression apparatus described above, the decompression apparatus can be realized by running a program that causes the computer to function as each unit described later on the computer.

  The decompression apparatus has an input unit 18 for inputting compressed data, and the input data is decoded by the decoding means 19 (step Sd1). In the decoding process, the above-mentioned compressed file is divided into divided image blocks 5 in a compressed state by block end marks 16, and each divided image block 5 is divided into sub-blocks based on sub-block end marks 14, and then sub-blocks. A step of decoding the compressed data in the block.

  When the decoding step is completed, the compressed file is in a state in which the format generation step (Sc4) in the compression process is completed, that is, information on the feature pixel 1 and the auxiliary pixel 3 for each color plane for each divided image block 5 in a predetermined format. After being described, the feature image restoration means 20 executes image generation (Sd2) using the feature pixel 1.

  In the feature image restoration step (Sd2), first, the feature pixel block 2, that is, the SB (ΔX) and SB (ΔD) sub-blocks are read, and the coordinate value of each feature pixel 1 on the bitmap space is determined from these. Calculated.

  On the other hand, the decoded auxiliary pixel 3 is embedded between the feature pixels 1 and 1 in the auxiliary image synthesis means 21 (step Sd3). In addition, the auxiliary image composition means 21 determines whether the adjacent feature pixels 1 and 1 are adjacent when the designation of the compression mode described in the block header 8 is a use value of the auxiliary pixel 3 or a compression value in the irreversible mode. Alternatively, the information missing dots are supplemented by interpolating between the auxiliary pixels 3, 3.

  The color planes generated by the auxiliary image synthesizing unit 21 are synthesized by the divided image block restoring unit 22 (step Sd4), and then synthesized by the divided image block synthesizing unit 23 based on the position information of the block header 8 and the like. (Step Sd5) and output from the output unit 24 as a restored image.

FIG. 2A is a functional block diagram of an image compression apparatus, and FIG. 2B is a functional block diagram of an image restoration apparatus. 1A and 1B are flowcharts showing an image compression method, and FIG. 1B is a flowchart showing an image restoration method. It is a flowchart which shows the detail of a feature image extraction step. It is explanatory drawing which shows the extraction method of the feature image. FIGS. 4A and 4B are diagrams illustrating pixel information conversion of auxiliary pixels, in which FIG. 4A illustrates a case where a difference from an adjacent pixel value is used as a conversion value, and FIG. 4B illustrates a pixel value difference from a feature pixel as a conversion value. It is a figure which shows a case. FIG. 5A is a diagram illustrating pixel information conversion of feature pixels, where FIG. 5A is a diagram illustrating a case where a pixel interval and a pixel value difference between adjacent feature pixels are used as a conversion value, and FIG. It is a figure which shows the case where an angle is used as a conversion value. It is a figure which shows the format of a compression file.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Feature pixel 2 Feature pixel block 3 Auxiliary pixel 4 Auxiliary pixel block 5 Divided image block

Claims (6)

A method of compressing image data in which a digital image is grouped and compressed by a pixel value difference with an adjacent pixel ,
Dividing the image into a predetermined number of color planes;
Each color plane
For a certain search direction, only a pixel whose pixel value difference with an adjacent pixel exceeds a predetermined threshold is extracted as a feature pixel, and a feature pixel block including a pixel group of only the feature pixel;
The remaining pixels associated with each feature pixel are extracted as auxiliary pixels and grouped into auxiliary pixel blocks consisting of a pixel group of only the auxiliary pixels;
A compression method of image data in which pixel information is converted and encoded in units of blocks. Wherein when the auxiliary pixel are continuous predetermined length to the search direction, image data compression method of claim 1, wherein extracting the final auxiliary pixels forming a predetermined length continuously in the search direction, wherein the pixel. The image data compression method according to claim 1 or 2, wherein the pixel information of the feature pixel is described by a converted value into relative position information from another immediately preceding feature pixel. The image data compression method according to claim 1, 2 or 3, wherein the pixel information of the auxiliary pixel is described by a conversion value into a difference from a corresponding characteristic pixel or a pixel value of another auxiliary pixel immediately before.   When the pixel value of the feature pixel belongs to another pixel column parallel to the search direction, and the distance between the pixels matches the pixel value of another feature pixel that is equal to or less than a predetermined threshold, the feature pixel The image data compression method according to claim 3 or 4, wherein the pixel value is described by a converted value into flag information indicating the presence of a feature pixel referred to in another column. After dividing the image into a plurality of image blocks,
Each divided image block is compressed by the method according to claim 1 or 2,
Thereafter, a compression method of image data to be combined and encoded by adding image block content information to the head of each divided image block,
The image block content information includes the number of each divided image block, the position in the image, the size information, and the compression mode of the auxiliary pixel block,
The compression mode is a compression method of image data that is set to either compression in the lossless mode of the auxiliary pixel block, nonuse, or compression in the lossy mode .

Images