JPS6324761A - Method of encoding image blocks - Google Patents

Abstract

(57) [Abstract] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION "Field of Industrial Application" The present invention relates to an image block encoding method for dividing an image into a plurality of blocks and encoding an image signal with redundancy suppression.

"Prior Art" Block encoding divides an image into multiple blocks, and for each block, each pixel in the block (3
In a method that calculates the average value of pixel signal values and encodes that average value as the representative value of the block, the conventional method For example, an adaptive block coding method was used, as shown on page 319 of the 1981 Television Society National Conference Lecture Proceedings.

That is, for a block of a predetermined size, the average value of each pixel in each block is calculated, and the signal average value (high level average value) of only the pixels whose level is higher than this average value within the block is calculated. , the signal average value of only pixels with small levels (
(low level average value). For each pixel in the block, the signal value is replaced with either the high level average value or the low level average value, and then compared with the true value. At this time, the root mean square error between the true value and the encoded value is calculated. If this root mean square error is larger than a predetermined threshold, the block is divided into four equal parts, and each of the small blocks is subjected to block encoding in the same way. If the root mean square error is smaller than a predetermined threshold, compare the high-level average value and the low-level average value, and if the difference is smaller than a separately predetermined threshold, use the intra-block average value. If the value is larger than a predetermined threshold value, the high-level average value, the low-level average value, and information indicating which of the average values is applied to each pixel are used as the code for that block. Therefore, in the adaptive block encoding method, it is necessary to frequently calculate the root mean square error, which inevitably increases the encoding time.

In addition, in the adaptive block encoding method, if there is little variation in pixel signal values within a block, the average value within the block is used to display all pixels within the block. Therefore, in a sloped region where pixel values change gradually, image quality deterioration in which block-like steps in signal values are observed in the reproduced image is likely to occur. In order to remove this, filter processing is required on the playback side, which requires memory and extra processing time.

Still natural image transmission is often applied to lines with a small transmission capacity, and therefore, if the amount of information is large, it takes time to transmit, which often causes discomfort to the user. In order to avoid this, it has been considered to show the rough contents of the image (W-child communication society journal 80/4 VOl, J63-B
N14 p, 379 9.386).

Such a transmission method has also been considered in the adaptive block encoding method, but in this case, blocks of various sizes are used, so it is necessary to arrange the transmission order after all the encoding is completed. be. When storing the encoding results in a storage device and transmitting the contents in response to a user's request, the time for adjusting the transmission order is not added to the transmission time, but the input image during communication , this adjustment time is added to the transmission time.

An object of the present invention is to provide an image block encoding method that reduces the amount of processing required for block encoding and eliminates the need for filter processing during decoding and reproduction.

"Means for Solving the Problems" This invention uses fixed-sized blocks and classifies each block into three types according to the amount of pixel value variation within the block.
(1) If the amount of variation is small, use the pixel signal value at the upper left corner of the block as the representative signal; <11) If the amount of variation is medium, use the high-level average value and low-level average value of the block. , the one close to the pixel signal in the upper left corner is taken as the representative signal, and the other is the average value, and binary information indicating whether each pixel in the block belongs to its high level average value or low level average value. (iii) If the amount of variation is large, the pixel signal at the upper left corner of the block is used as the representative signal, and all remaining pixel signals are used.

In this way, determining which of the three each block belongs to is not determined based on the result of calculation using the true value and the encoded value, but is determined based on the characteristics of the true value itself, thereby significantly reducing the amount of processing (1).

In the case of a block with a small amount of variation on the decoding side, filter processing can be performed simply by PR by interpolating and calculating intra-block pixel signals using representative signals of surrounding blocks.

According to the present invention, an image is divided into blocks of, for example, N×N pixels, a high level average value and a low level average value are determined for each block in the same way as in adaptive block coding, and the difference between the two average values is determined, Based on the difference value, the encoding method of the block is determined as described above (i) (ii
) (iii) is determined. That is, the encoding method for each block is determined by the human pixel signal value.

The encoding method for each block is as follows: (i) NxN pixels are represented by the signal value of the pixel in the upper left corner, (ii) NxN pixels are represented by the high level average value, low level average value, and each pixel in the block is represented by the signal value of the pixel in the upper left corner. Either quantization is performed into a high level average value or a low level average value, or (iii) the signal value of N×N pixels is used as is.

When decoding the code block (i) on the decoding side, the representative signals of the blocks to the right, directly below, and diagonally below the right are used to obtain the remaining (NXN-1)ii! of that block. A filter effect can be obtained by calculating the signal values of the ii elements, but if there is a sharp edge in the image between adjacent blocks, using the signal values of the adjacent blocks for decoding will cause significant deterioration. In order to avoid this, the coding method (11) or (iii) can be used for blocks that are likely to suffer from this.

Preferably, the code transmission order is such that for each block, the identification signal indicating the type of encoding and the representative signal of that block are first sent to the block configuration class, that is, first the identification signal and representative signal of the block in the upper left corner of the image, and then the identification signal and representative signal of the block in the upper left corner of the image. The identification signal and representative signal of the block on the right are sequentially sent from the upper left corner to the lower right corner corresponding to each block, and then (11
The remaining average value and quantization distribution information for the block encoded in (iii) are sent out, and finally the pixel signals for the remaining (NXN-1) pixels encoded in (iii) are sent out.

Embodiment FIG. 1 is a diagram illustrating the encoding method of the present invention and its code transmission method, in the case where each block is made up of 4 pixels×4 pixels. On the encoding side, each pixel value of the image is stored in image memory 1.
1 (part of which is shown in Figure 1), and the memory contents are shown in the figure for four blocks B, BS, Bs, B, corresponding to the arrangement of blocks in the image, and each Each pixel 12 corresponds to its pixel arrangement within the block.
Each signal value is shown as a numerical value in a circle. Block B1 is an example in which only the average value is output, and therefore encoding method (i) is performed using the average value as the representative signal, and blocks Bz and B4
Block B is an example of performing encoding method (ii) that outputs the high-level average value, low-level average value, and their distribution, and block B is an example of performing encoding method (iii) that outputs the signal values of all pixels. be. Each block Bl.

B2. B3. Each identification signal 13 and representative signal 1 for B4
Indicates a block code C,, cz, C3°C4 consisting of blocks Bz, B4 . B,! , correction code Sz
, Sa, and Sz are added, respectively.

The sending order of these codes is CI...C2...C3...
C4, and after outputting all block codes, the correction code is B2
...S, ...Sj.

Each pro/7ri is processed in order from the block in the upper left corner of the image.
The internal pixel value variation amount is checked and the corresponding representative signal 14 and its identification signal 13 are output as illustrated by symbols C9 to C4. Taking block B2 as an example, the average value of the pixel values within the block is 33.9, the high level average value is 38, the low level average value is 32, and the difference between the high level average value and the low level average value is 6. If the small threshold TL is 1 and the large threshold Th is 7, this block is encoded using encoding method (ii).
is applied, and the pixel in the upper left corner is low and healthy, so
The low level average value 32 is output as the representative signal 14, and 1 is output as the identification signal 13 indicating the encoding method (11). In the case of block B, the intra-block average value is 27.5, the high level average value is 42, and the low level average value is 19, and the difference is greater than the large threshold value 7, so all pixel values are The identification signal 13 indicating the encoding method (iii) to be output is 2, and the representative signal 14 is the value 15 in the upper left corner.
Output.

After all block codes made up of each identification signal 13 and representative signal 14 are output, block Bz is next output.

The correction code for the encoded block by the encoding method (ii) represented by B, is output as illustrated in S,,S,. In the case of block B2, the other average value (No. 3 15, in this example, a signal indicating a high level average value 38) that is not used as the representative signal of block code C2, and the signal indicating which average value each pixel belongs to (16) are used. Biff) binary signal 16 is output. In this example, the binary signals 16 are expressed in order from the upper left corner to the lower right corner, with 1 indicating that the attached average value signal 15 is used and O indicating that it is not used. Identification signal 1
After 3 finishes outputting the correction code for 1, block B
The correction code for the all-pixel transmission block represented by , is output as illustrated in S. In this example, the correction code S outputs pixel value signals in order from the upper left corner to the lower right corner of the block.

FIG. 2 shows the notation tα of part 17 of the image memory on the decoding side.
The state is shown in correspondence with the image, and the block B+ in FIG.
,B2. Ba, blocks DB+, D corresponding to Ba
The values of each pixel of Bz, DBz, and DB4 are shown. These blocks DBI.

The values of each pixel in D B2 , D B3 , and D B4 are block codes C,, cz, C3, and C4, respectively.
This is the value decoded based on.

On the decoding side, the manual code C8...C2...C1...
Each pixel value in each block is determined using the representative signal 14 in the order of C4, that is, starting from the top left corner. On the decoding side, the representative signal 14 of each block is stored, as well as the identification signal 13 of each block. When determining the pixel value in each block, the identification signal 13 is 01 block DB.
1 and this block D B 3 are determined by interpolating the signal value of each pixel together with the representative signal 14 of the adjacent block.

Therefore, for these blocks DB, DB3, it is necessary to wait for the input of each adjacent block representative signal 14 to determine the pixel values. Therefore, on the decoding side, after inputting block representative signals for two rows, display processing is started.

For example, interpolation of pixel values for blocks whose identification signal 13 is 0 or 2, such as blocks DB+ and DB*, is as follows. Now, as shown in Fig. 3, the pixel value of each pixel in one block DB is P, ~P44, the value of the representative signal of the block to the right of this block is P, and the value of the representative signal of the adjacent block below. Assuming that pus is the value of the representative signal of the lower right adjacent block and pss is the value of the representative signal of the lower right adjacent block, each pixel value in the bronc DB is obtained by the calculation of the following equation.

P+z= (3P++ + Pu5) /4P 13=
(P ++ + P +s) /2P14= (P1
1+ 3 Pu5) /4Pz+= (a Pz10Ps
+) /4P! 2= (3P12+ (3PSI +P
S5) /4)/4P2 co=(3P13+(PSI+P
SS) /2)/4Px4= (3PI4+ (PSl
+ 3 Psi) /4)/4P11= (pH”PS
I> /2 Psz= (P1i+ (3Ps++Pss) /4)
/2Pxs=CPIs+ (P51+P5S)/2)
/2P34= <r'+4+ (Ps+ + 3 Ps
s) /4)/2Pa+= (Pz+3 Ps+ll
/4P42= (Puz + 3 (3Ps++Ps
s) /4)/4P43-(P +s + 3 (P
s+”P ss) /2)/4P aa= (P
la+ 3 (P s++3 P) /4)/4 On the other hand, the identification signal 13 is 1 like blocks DBZ and DB.
For the block, all pixels in the block are taken as its representative signal value.

In other words, when decoding and reproducing the image, the block code...
・As soon as C,...C2...Cs...C4... are obtained and the interpolation becomes possible, the pixel values in each block are determined according to 0 to 2 of each identification signal. Each pixel value is determined and displayed using the obtained pixel value. At this time, if the identification signal is O, the interpolation makes the border between blocks less noticeable, and if the identification signal is 2, the pixel values within the original block have fluctuated greatly, so if the fluctuations are also large for the surroundings. considered,
If all pixel values are displayed using only one of the representative signals 14, they will be greatly different from those of the ring block, but it is possible to reduce the difference as a whole by the interpolation. When the identification signal is 1, there is not much variation, so all pixels are displayed with the representative signal value.

When the reproduction of each pixel value using only the block representative signal has been completed for all blocks, the correction code S is input. In advance,
Since the identification signal 13 is obtained for each block, the correction code S
There is no need to add block identification information to each correction code S as long as the input order is determined.

An example of each pixel value corrected by the correction code S is stored in the professional DB.
+, DBz, DB3, DB4, blocks HBI, H in a part of the image memory 18 for decoding are shown in FIG.
B! , HB□, HB4.

For blocks whose identification signal 13 is O, such as block D B l , no correction is performed, and pixel values obtained only from representative signal values are used. Therefore, block HB is block D B l
completely equal to For a block in which the identification signal 13 is 1, for example DBz, its correction code S2 is used, and the pixel value of the pixel corresponding to 1 in the 2 plant code 16 (block D
B1) is corrected to the value 38 of the average value signal 15 of the correction code S2 to obtain a corrected block HB2. Similarly, for block DB4, 2 in correction code S4 is
The pixel value 46 of the pixel corresponding to 1 in the plant code 16 is corrected to 40 to obtain a corrected block HB4. After the block with the identification signal 13 of 1 is corrected in this way, the block with the identification signal 13 of 2 is corrected. In this correction, for example, for block DB3, each pixel value is corrected to the signal value of the correction code S in order from the upper left to the lower right, except for the pixel at the upper left corner of the block, using the correction code S. Obtain block HB.

As shown in FIG. 5, a part of the image memo IJl l storing the original image, blocks BS, B? The amount of variation in pixel values within that block is smaller than the small threshold value T , and the blocks B6 , Bs , B6 . B+1 means that the amount of variation in pixel values within the block is larger than the threshold value Th, and the professional
Block B adjacent to the lower side of B7. The amount of pixel value variation within the block is between the threshold value Tj and Th, that is, the blocks Bs and Bt where the pixel value variation within the block is small are located near the image boundaries (blocks Bb, Ba to B11).
may exist.

In such a case, when decoding is performed as explained with reference to FIGS. 2 and 3, each pixel value of blocks B, ~B, becomes blocks HB, ~HBI, respectively, in FIG. 6.

It becomes as shown inside. Blocks BS and B with small fluctuations
Since the decoding of T is performed by interpolation using representative signal values of surrounding blocks, some of the pixels of blocks HBs and HBy have a large difference from the original pixel value. For example, block HBi
Pixel values such as 19.2124 within are original pixel values of 15, 14.
It is too different from 16 etc. This is because the interpolation acts to eliminate large changes in the boundaries of the image. Therefore, if there is a block near the boundary of the image that only undergoes interpolation and no correction during decoding, that is, a block with an identification signal of 0, this is checked in advance on the encoding side, and the two The average value and its distribution may be encoded using two system codes, that is, the identification signal 13 may be encoded as a block of 1.

In this way, the following procedure can be used to check whether a block with the identification signal 13 of 0 exists near the boundary. That is, it is checked whether there is a block with the identification signal 13 of 2, which outputs all pixel values as shown in FIG. It is determined that the block is located near the boundary of the image, and the block is encoded in the same way as the block whose identification signal is 1. A block with a small amount of variation in pixel values within the block may be located near the boundary of an image even if it is not adjacent to a block with a large amount of variation. That is, for example, as shown in FIG. 7, which shows a part of the memory contents of the original image, the fluctuations of the pixel values within each block Bs and Bb are smaller than a predetermined value, and only the pixel value in the upper left corner is used as the representative signal. 1
4, and the surrounding blocks Bt
.

Be, Baa, and Bz are blocks in which the identification signal is 1 and the amount of variation in pixel value is intermediate, but if only the representative signal value is output for block B6 and interpolated with the representative signals of the surrounding blocks on the decoding side, as shown in FIG. In particular, the right adjacent block B7, the upper adjacent block B1°, and the lower right adjacent block B
ll, resulting in a value considerably different from the original pixel. In this case, the difference between the representative signal value 35 of block B and the representative signal value 27 of the right adjacent block B is 35-27=8
is larger than a predetermined threshold, for example 4, the representative signal value 27 of the right adjacent block B and the block B
If it is smaller than 1, that is, 4, of the further right neighbor in the arrangement direction of a and B, this block B is determined to be near the boundary of the image, and this block B uses two representative signal values, and It is assumed that the identification signal outputted from the distribution No. 2 plant number 16 is also encoded as 1. Regarding block B, other directions sequentially arranged adjacent to this, for example, block B4,
When the same determination as above is made using each representative signal value of B++, 815, in this example, block B is similarly determined to be near the boundary, but in such a determination, the block B is determined to be in the rightward direction, downward direction, If any one of the blocks in the lower right direction is determined to be near the boundary, that block B is encoded by method (ii) or method (iii).

Note that when the above determination is made for block B, the difference between its representative signal value 37 and the representative signal value 32 of the adjacent block E3+o is 5, which is larger than the predetermined threshold value 4. , the difference between the representative signal value 32 of block B1° and the representative signal value 29 of block B I5 which is further adjacent in the arrangement direction of blocks Bs and B+o is 3,
Since this value is larger than half of the difference, that is, 5/2, it is determined that this block B is in a part where the pixel value changes are sloped and is not near the boundary, and this block BS
is encoded using encoding method (i). In this manner, the pixel value distribution shown in FIG. 6 is block coded, and the decoded state is shown in FIG. It will be understood that this is relatively close to the original image.

For example, when transmitting an image that changes relatively slowly and is close to a still image, such as in a video conference, over a line with a small transmission capacity, first all block codes consisting of the identification signal 13 and the representative signal 14 are transmitted as in the above example. If you transmit about the screen, the receiving side can reproduce the approximate image with only this,
The subsequently received correction code allows the image to be corrected to a fairly good quality. Furthermore, in this case, as described above, on the receiving side, by interpolating using representative signals of peripheral blocks, filter processing efficiency in which block boundaries are not noticeable can be obtained without performing filter processing that requires a large amount of processing.

This invention is not limited to application to compressing image signals when transmitted in this way and decoded and reproduced on the receiving side, but can also be applied when storing many images as files. If it is not necessary to obtain the identification signal, the representative signal, and the correction code for each block if the identification signal is 1 or 2, it may be arranged in the order of the blocks as one code.

In the above, when there is a block in which the identification signal 13 is 0 near the boundary of the image, that block is encoded using the encoding method (ii) in which the identification signal 13 is 1. Encoding may be performed using this encoding method (iii). In addition, such identification signal 1 near the boundary
Without changing the encoding for the block where 3 is 0,
Encoding may be performed using encoding method (i).

``Effects of the Invention'' As stated above, this invention uses a fixed block method that does not change the size of the blocks, and uses addition, subtraction, and division to perform encoding methods (i), (ii), and (iii).
The amount of processing required for encoding is smaller than that of adaptive block encoding. If interpolation is performed using representative signal values of adjacent blocks during decoding, less decoding processing is required and a filter effect can be obtained. Another advantage is that since fixed blocks are used, no special processing is required to allow the decoding side to recognize the entire image at an early stage of code transmission. In addition, by making the threshold value that determines the encoding method variable on the encoding side, the encoded image quality can be controlled on the encoding side, and the decoding side can maintain the same IS without considering changing the threshold value. It can be decrypted using this method.

[Brief explanation of drawings]

Figure 1 is a diagram showing an example of a part of the storage contents of a memory that stores pixels of an original image on the encoding side and an example of encoding each block thereof, and Figure 2 is a diagram showing an example of the blocks in Figure 1 on the decoding side. Fig. 3 shows an example of the memory contents in Fig. 2 corrected by a correction code, Fig. 4 shows interpolation using representative signals of adjacent blocks. A diagram showing pixel values in one block and representative signal values of adjacent blocks used to explain the calculation, FIG. 5 is a diagram showing an example of a block with an identification signal of 0 near the image boundary,
Figure 6 shows the state of interpolation decoding for the blocks in Figure 5, and Figure 7 shows another example of the image memory for an image in which blocks with small pixel value fluctuations are near the boundary. Figure 8 is a diagram showing each pixel value obtained by encoding only one representative signal value for this block B, and interpolating and decoding this with the representative signal values of surrounding blocks, and Figure 9 is a diagram shown in Figure 7. FIG. 12 is a diagram showing a storage state of a decoded image when a block B having a small intra-block pixel value variation near the image boundary is encoded by encoding method (ii).

Claims (1)

[Claims] (1) When an image is divided into multiple blocks and the amount of variation in the image signal value of each pixel in each block is smaller than the smaller of two predetermined thresholds, large and small, and If the value is greater than the threshold value, the pixel signal value at the upper left corner of the block is set as the block representative signal, an identification signal indicating what kind of variation the block has is added, and the image signal value in the block is If the amount of variation is smaller than the first large threshold and larger than the small threshold, high-level averaging is performed that targets only pixels that have a signal value larger than the signal average value in that block. A low level average value is calculated for only pixels that have a signal value smaller than the above average value, and if the pixel in the upper left corner of the block is larger than the average value, a high level average value is determined. In this case, the low level average value is used as the block representative signal and its identification signal is attached. A binary signal indicating the high level or low level average value of the transmission and which average value each pixel in the block has is added as a correction code, and the block whose variation amount is larger than the threshold value is added with that value. An image block encoding method that adds the signal values of all remaining pixels, excluding the pixel value at the upper left corner of the block, as a correction code.