JP2004023370A - Image encoding method and apparatus - Google Patents

Abstract

Kind Code: A1 The present invention significantly reduces block distortion and quantization distortion between mutually adjacent pixels, and encodes a high-quality and high-compression image without requiring special data.
An image signal is input to an input unit, and an 8 × 8 block input circuit 2a divides the image into blocks of 8 × 8 pixels and detects vertices. The prediction error quantizer 3a calculates and quantizes the prediction error of the detected vertex. The prediction error inverse quantizer and new vertex calculation circuit 4a inversely quantizes the error quantized by the prediction error quantizer 3a to newly obtain a prediction error, and calculates a new vertex. The new pixel calculation circuit 5a calculates a new pixel in the block using the new vertex and the bilinear surface. The error calculation circuit 6a compares the new pixel calculated by the new pixel calculation circuit 5a with 64 pixels in the block to calculate an error. Each of the error comparison circuits 7a compares the error calculated by the error calculation circuit 6a with a threshold value.
[Selection diagram] FIG.

Description

[0001]
TECHNICAL FIELD OF THE INVENTION
The present invention relates to an image encoding method and apparatus for encoding and processing image signals for transmission and storage, which are applied to videophones, surveillance cameras, and the like.
[0002]
[Prior art]
As a general compression standard for image signals, JPEG has been proposed for still images, and MPEG has been proposed for moving images. Typical image compression methods include predictive coding and vector quantization as non-block coding in addition to block coding used in JPEG and MPEG.
[0003]
In the case of MPEG, although the compression ratio is high, mosquito noise and block distortion occur because the data is divided into blocks and the DCT coefficients are reduced.
[0004]
Predictive coding is coding in which the pixel value is predicted from pixels adjacent to each other and the prediction error is quantized. In this case, although mosquito noise and block distortion do not occur, quantization distortion increases as the compression ratio increases. The deterioration of the image quality due to the above becomes remarkable.
[0005]
Vector quantization is a method of replacing a plurality of pixel values with a predetermined representative vector, and can increase the compression ratio as compared with the case of predictive coding. In such a case, it is difficult to generate a codebook (encoded data, that is, a table). If the type of the codebook is reduced, block distortion is likely to occur.
[0006]
[Problems to be solved by the invention]
An object of the present invention is to significantly reduce block distortion and quantization distortion between pixels adjacent to each other, and perform high-quality and high-compression-rate image encoding without requiring special data. An object of the present invention is to provide an image encoding method and an apparatus therefor.
[0007]
[Means for Solving the Problems]
The image encoding method according to the present invention comprises:
The entire screen is divided into predetermined pixel units,
Using the vertices and bilinear surfaces of the pixels in the block, calculate a new pixel value,
Calculate the difference between the pixel value in the block and the new pixel value,
Based on whether or not these differences are within a predetermined range, determine whether to encode the block or to further divide the block,
It is characterized in that the encoding process is performed in an appropriate pixel unit block according to the result of the determination.
[0008]
According to the present invention, a decision as to whether to encode a block divided in a predetermined pixel unit or to further divide a block is made using a new pixel value, and an appropriate pixel according to the result of the decision is made. By performing the encoding process on the unit block, problems unique to MPEG, predictive encoding, and vector quantization do not occur. As a result, block distortion and quantization distortion between pixels adjacent to each other are significantly reduced, and high-quality and high-compression-rate images can be encoded without requiring special data.
[0009]
Preferably, a histogram in which the maximum value of the absolute value of the difference is obtained for each block is created, and the predetermined value is calculated from the histogram.
[0010]
An image encoding device according to the present invention includes:
An image encoding device for encoding and transmitting an image signal,
Block dividing means for dividing the entire screen into blocks of a predetermined pixel unit,
New pixel value calculation means for calculating a new pixel value using the vertices and bilinear surfaces of the pixels in the block,
A pixel value in the block, and a difference calculation unit for calculating a difference between the new pixel value,
A determination unit configured to determine whether to encode the block or to further divide the block, based on whether these differences are within a predetermined range;
Encoding means for performing an encoding process on an appropriate pixel unit block according to the result of the determination.
[0011]
ADVANTAGE OF THE INVENTION According to this invention, block distortion and quantization distortion between mutually adjacent pixels are significantly reduced, and high-quality and high-compression-rate images can be encoded without requiring special data. .
[0012]
BEST MODE FOR CARRYING OUT THE INVENTION
To facilitate understanding of the present invention, first, the basic principle of the present invention will be described.
Image compression methods fall into two categories with respect to the use of correlation.
The first method is a method using the correlation between mutually adjacent pixels. Specifically, the first method is to reduce the number of quantization bits by using the difference between mutually adjacent pixels as in predictive coding. For example, a method of performing spatial processing or a method of performing frequency processing such as DCT or wavelet transform using a transform system to reduce high-frequency coefficients.
However, in the first method, there is a possibility that the quantization distortion becomes remarkable or the block distortion or mosquito noise occurs as the compression ratio is increased.
[0013]
The second method is a method of calculating a correlation between a predetermined pattern and an image and approximating the image to a pattern having a high correlation. Specific examples include fractal coding and surface approximation. The pattern is obtained by extracting from the image or calculating from an appropriate relational expression.
However, even if a coefficient is obtained by calculation or a method such as matching or correlation is used, it is actually difficult to detect an optimum pattern.
[0014]
The present invention uses curved surface approximation. However, for a high-order curved surface, it is not easy to determine the shape, order, coefficient, and the like of the curved surface approximating the image by a simple method. Therefore, although an image is approximated using a bilinear surface, an image region is divided for an image having a complex order. Specifically, a relatively large rectangular area is calculated using four vertices of the rectangular area, and an internal point is approximated. However, if the error from the calculated value is large, the rectangular area is divided and the calculation is performed using the four vertices of the divided rectangular area. Such division is repeated up to the minimum rectangular area, and if the division cannot be made even if the division is made up to the minimum rectangular area, the error from each pixel is quantized.
[0015]
According to the present invention, the bilinear surface is calculated based on a predetermined formula using the pixel values of the image, so that it is not necessary to use a table of the surface. Further, by forming the rectangular area with blocks and vertices as described later, it is possible to simplify the calculation of the internal points and to prevent the distortion of the block boundary.
[0016]
The process using the bilinear curved surface will be described in more detail. First, the entire screen is divided into blocks of a predetermined pixel unit. Each of these blocks can be, for example, a block of 8 × 8 pixels having 8 pixels in the horizontal direction and 8 pixels in the vertical direction, but the size of each block can be set arbitrarily.
[0017]
The block is further divided, for example, when divided into 8 × 8 pixel blocks, the blocks are sequentially divided into 4 × 4 pixel blocks and 2 × 2 pixel blocks. Here, when the vertex of the block is divided into blocks of, for example, 8 × 8 pixels, it corresponds to the pixel a at the upper left as shown in FIG. 1, but other pixels can be selected as the vertex. The same applies to a block of 4 × 4 pixels and a block of 2 × 2 pixels.
[0018]
That is, in the divided block of the largest unit, the strength of the correlation between the value calculated using the equation of the bilinear surface and the pixel value of the image is obtained, and if there is a correlation, the vertex of the block is determined. Use a representative value. If there is no correlation, the block is further divided, and the strength of the correlation is similarly obtained in the divided blocks. The block division is repeated until there is a correlation, but if it is determined that there is no correlation even if it is divided into the smallest block that can be divided, in addition to the vertices of the block, from the vertices and the bilinear surface, The error between the calculated new pixel and each pixel is quantized.
[0019]
The coordinates of the four corners of the bilinear curved surface Q (u, w) are defined as P (0,0), P (0,1), P (1,0), and P (1,1), respectively. The case is represented by the following equation.
[Equation 1]
Q (u, w) = P (0,0) (1-u) (1-w) + P (0,1) (1-u) w + P (1,0) u (1-w) + P (1, 1) uw
[0020]
According to the present invention, the pixels at the vertices of three adjacent blocks (right, lower, and diagonally lower right) are defined as P coordinates, and the value corresponding to u, w of the bilinear curved surface Q (u, w) is set to 1. By using a value that can be expressed by a power of 2 such as / 2, 1/4, 1/8, only addition / subtraction and shift operations are performed. For example, when 1/8 = 0.125, 3 bits are used. All you have to do is shift right.
[0021]
Next, an embodiment of an image encoding method and an image encoding apparatus according to the present invention will be described in detail with reference to the drawings.
FIG. 2 is a flowchart of the operation of the embodiment of the image encoding method according to the present invention. In this case, the entire screen is divided into 8 × 8 pixel blocks, and the thus divided 8 × 8 pixel blocks are used as a reference.
[0022]
The control routine illustrated in FIG. 2 is executed, for example, on the transmission side of the image monitoring system. First, in step S1, a threshold value for a determination process is calculated as preprocessing for encoding. Step S1 will be described later in detail with reference to FIGS.
[0023]
Next, in step S2, the number J of all blocks when the entire screen is divided into blocks of 8 × 8 pixels is set. The total number of blocks J is determined by the number of vertical and horizontal pixels. Next, in step S3, processing of a block of 8 × 8 pixels, which will be described later in detail with reference to FIGS. 7 and 8, is performed.
[0024]
Next, 1 is subtracted from the total number of blocks J (step S4), and thereafter, it is determined whether or not the processing of all the blocks is completed (step S5). If the processing has been completed, this routine ends. If not, the processing returns to step S3.
[0025]
FIG. 3 is a flowchart of the first half of the subroutine of step S1 in FIG. 2, and FIG. 4 is a flowchart of the second half of the subroutine of step S1 in FIG. In this embodiment, the reference block is composed of 8 × 8 pixel blocks. Since processing is performed on the basis of an 8 × 8 pixel block, the block is divided into 4 × 4 pixel blocks if it is divided in half vertically and horizontally, and further divided into 2 × 2 pixel blocks if it is further divided into half. Will be done. The subroutine of FIGS. 3 and 4 calculates the threshold value by proceeding with the calculation of the block of 4 × 4 pixels and the block of 8 × 8 pixels based on the block of 2 × 2 pixels.
[0026]
First, in step S11, the number m of blocks when the entire screen is divided into blocks of 2 × 2 pixels is set. Next, in step S12, a new pixel (new pixel) is calculated from the vertex of the block of 2 × 2 pixels and the vertices of three blocks adjacent thereto. For example, as shown in FIG. 5, from the pixels A0, A2, C0 and C2, a new pixel in a block having the pixel A0 as a vertex is calculated by a bilinear curved surface.
[0027]
Next, each pixel in the block (pixels A0, A1, B0, and B1 in FIG. 5) is compared with the new pixel calculated in step S12 (step S13), and the maximum absolute value of the compared difference is detected. Then, it is registered in a first histogram (histogram 1) described later (step S14), 1 is subtracted from the number m of blocks (step S15), and whether or not the number m of blocks is 0, that is, 2 × 2 It is determined whether the processing has been completed for all of the pixel blocks. If the process has been completed, the process proceeds to the next step S17 to perform the process on the block of 4 × 4 pixels, whereas if the process has not been completed, the process returns to step S12.
[0028]
Next, in step S17, the number of blocks 1 when the entire screen is divided into blocks of 4 × 4 pixels is set. Next, in step S18, a new pixel (new pixel) is calculated from the vertex of the block of 4 × 4 pixels and the vertices of three blocks adjacent thereto. For example, as shown in FIG. 5, a new pixel in a block having the pixel A0 as a vertex is calculated from the pixels A0, A4, E0, and E4 by a bilinear curved surface.
[0029]
Next, each pixel in the block (pixels A0 to A3, B0 to B3, C0 to C3, and D0 to D3 in FIG. 5) is compared with the new pixel calculated in step S18 (step S19), and the difference of the compared difference is determined. The maximum value of the absolute value is detected and registered in a second histogram (histogram 2) described later (step S20), and 1 is subtracted from the number l of blocks (step S21). No, that is, whether or not the processing has been completed for all of the 4 × 4 pixel blocks. If the process has been completed, the process proceeds to the next step S23 to perform the process on the block of 8 × 8 pixels, whereas if the process has not been completed, the process returns to step S18.
[0030]
Next, in step S23, the number k of blocks when the entire screen is divided into blocks of 8 × 8 pixels is set. Next, in step S24, a new pixel (new pixel) is calculated from the vertex of the block of 8 × 8 pixels and the vertices of three blocks adjacent thereto. For example, as shown in FIG. 5, a new pixel in a block having the pixel A0 as a vertex is calculated from the pixels A0, A8, I0, and I8 by a bilinear curved surface.
[0031]
Next, each pixel in the block (pixels A0 to A7, B0 to B7, C0 to C7, D0 to D7, E0 to E7, F0 to F7, G0 to G7, and H0 to H7 in FIG. 5) is calculated in step S24. (Step S25), and the maximum absolute value of the compared difference is detected and registered in a third histogram (histogram 3) described later (Step S26). Is subtracted (step S27), and it is determined whether or not the number of blocks k is 0, that is, whether or not the processing has been completed for all the blocks of 8 × 8 pixels. When the processing is completed, the threshold values for the block of 2 × 2 pixels, the block for the block of 4 × 4 pixels, and the block for the block of 8 × 8 pixels and the histograms 1, 2, and 3 as shown in FIG. Three types of quantization tables for each vertex of the 2 × 2 pixel block, the 4 × 4 pixel block, and the 8 × 8 pixel block are determined (step S29), and this routine ends. On the other hand, if the processing has not been completed, the process returns to step S24.
[0032]
FIG. 7 is a flowchart of the first half of the subroutine of step S3 in FIG. 2, and FIG. 8 is a flowchart of the second half of the subroutine of step S3 in FIG. This subroutine performs variable length coding.
[0033]
First, in step S31, a prediction error of a vertex of a predetermined block of 8 × 8 pixels is quantized based on the quantization table calculated in step S29 of FIG. For example, when a block is divided as shown in FIG. 9A, pixels A0, A8,. . . I0, I8,. . . Is quantized.
[0034]
Next, in step S32, the quantized prediction error is inversely quantized to calculate the error, and the original vertex is corrected by adding and subtracting it to calculate a new vertex. In the case of FIG. 9A, as shown in FIG. 9B, A0 ′, A8 ′, I0 ′ and I8 ′ are new vertices.
[0035]
Next, a new pixel is calculated from the new vertex and the bilinear curved surface (step S33), and the new pixel calculated in this way and pixels in the block (for example, 64 pixels A0 to A7, B0 to B7, 64 pixels in FIG. 9A). C0 to C7, D0 to D7, E0 to E7, F0 to F7, G0 to G7, and H0 to H7) (step S34).
[0036]
Next, in step S35, as a result of the comparison, it is determined whether or not the maximum value of the error is within the threshold value for the block of 8 × 8 pixels obtained in step S29 of FIG. If the value is within the threshold value, the subroutine ends to perform the processing for the next block of 8 × 8 pixels, whereas if the value is outside the threshold value, the processing of the block of 4 × 4 pixels is performed. Proceed to step S36 to perform.
[0037]
In step S36, 4 is set as the number k of blocks of 4 × 4 pixels. Next, in step S37, among the vertices of the predetermined 4 × 4 pixel block, the prediction errors of vertices other than the vertices quantized in step S31 are quantized based on the quantization table calculated in step S29 in FIG. I do. For example, when the block is divided as shown in FIG. 9A, the prediction errors of the pixels A4, E0, E4, E8, and I4 are quantized.
[0038]
Next, in step S38, the quantized prediction error is inversely quantized to calculate the error, and correction is performed by adding and subtracting the original vertex to calculate a new vertex. In the case of FIG. 9A, as shown in FIG. 9C, A0 ', A4', A8 ', E0', E4 ', E8', I0 ', I4' and I8 'are new vertices.
[0039]
Next, a new pixel is calculated from the new vertex (in the first block, A0 ′, A4 ′, E0 ′ and E4 ′ as shown in FIG. 9D) and the bilinear surface (step S39), and the new pixel thus calculated is calculated. And the pixels in the block (for example, 16 pixels A0 to A3, B0 to B3, C0 to C3, and D0 to D3 in FIG. 9A) (Step S40).
[0040]
Next, in step S41, as a result of the comparison, it is determined whether or not the maximum value of the error is within the threshold value for the block of 4 × 4 pixels obtained in step S29 of FIG. If it is within the threshold value, the process proceeds to step S51 described below to perform processing on the next block of 4 × 4 pixels, whereas if it is outside the threshold value, the block of 2 × 2 pixels is processed. The process proceeds to step S42 to perform the process of.
[0041]
In step S42, 4 is set as the number 1 of blocks of 2 × 2 pixel blocks. Next, in step S43, among the vertices of the predetermined 2 × 2 pixel block, the prediction errors of the vertices other than the vertices quantized in steps S31 and S37 are calculated based on the quantization table calculated in step S29 in FIG. And quantize. For example, when the block is divided as shown in FIG. 9A, the prediction errors of the pixels A2, C0, C2, C4 and E2 are quantized.
[0042]
Next, in step S44, the quantized prediction error is inversely quantized to calculate the error, and the original vertex is corrected by adding and subtracting it to calculate a new vertex. In the case of FIG. 9A, as shown in FIG. 9E, A0 ', A2', A4 ', C0', C2 ', C4', E0 ', E2', and E4 'are new vertices.
[0043]
Next, a new pixel is calculated from the new vertex (A0 ′, A2 ′, C0 ′ and C2 ′ as shown in FIG. 9F in the first block) and the bilinear surface (step S45). And the pixels in the block (for example, four pixels A0, A1, B0 and B1 in FIG. 9A) are compared (step S46).
[0044]
Next, in step S47, as a result of the comparison, it is determined whether or not the maximum value of the error is within the threshold value for the block of 2 × 2 pixels obtained in step S29 of FIG. If it is within the threshold value, the process proceeds to step S49 to be described later in order to perform processing on the next block of 2 × 2 pixels. After quantizing the difference from the pixel for each of the four pixels (step S48), the process proceeds to step S49.
[0045]
In step S49, 1 is subtracted from the number l of blocks. Thereafter, in step S50, it is determined whether or not the processing has been completed for all the blocks of 2 × 2 pixels. When the process is completed, the process proceeds to a step S51, whereas when the process is not completed, the process returns to the step S43.
[0046]
In step S51, 1 is subtracted from the number k of blocks. Thereafter, in step S52, it is determined whether or not the processing has been completed for all the blocks of 4 × 4 pixels. When the processing is completed, this subroutine is completed. On the other hand, when the processing is not completed, the processing returns to step S37.
[0047]
The subroutine shown in FIG. 7 and FIG. 8 is performed for all of the divided blocks of 8 × 8 pixels.
Due to the quantization error of the value of each vertex, the coded bit generation amount predicted in the initial stage fluctuates. Therefore, for each block line of 8 × 8 pixels, referring to the histograms of the block of 8 × 8 pixels, the block of 4 × 4 pixels, and the block of 2 × 2 pixels, the number of blocks predicted to be within the threshold value and When the number of blocks within the threshold is relatively small, that is, when the number of encoded bits is relatively large, the comparison is performed with the number of blocks within the threshold in the actual processing. Increase the threshold to reduce the amount of generated coding.
[0048]
Note that the image decoding process on the receiving side of the image monitoring system is performed by performing an operation reverse to the flowchart of FIG.
[0049]
10 to 13 are flowcharts of the operation of another embodiment of the image encoding method according to the present invention.
First, in step S101, the number J of all blocks when the entire screen is divided into blocks of 8 × 8 pixels is set.
[0050]
Next, in step S102, four vertices of the block of 8 × 8 pixels are quantized. For example, pixels A0, A8, I0 and I8 in FIG. 14A correspond to four vertices. Next, in step S103, the vertex quantized in step S102 is inversely quantized to calculate a new vertex. For example, new vertices of four vertices corresponding to pixels A0, A8, I0, and I8 in FIG. 14A are pixels A0 ', A8', I0 ', and I8', respectively.
[0051]
Next, in step S104, a new pixel in a block of 8 × 8 pixels is calculated from the four new vertices calculated in step S103 and the bilinear surface, and a residual between the new pixel and each pixel is obtained. . These residuals are represented by A0'-1, A1-1,. . . H7-1 (FIG. 14B).
[0052]
Next, in step S105, the maximum value (absolute value) of the residual between the new pixel and each pixel is detected and stored in the first residual table (residual table 1). Register in the histogram (Histogram 1).
[0053]
Next, 1 is subtracted from the total block number J (step S106), and thereafter, it is determined whether or not the processing for all blocks, that is, the processing for the entire image (one screen) is completed (step S107). When the process is completed, the process proceeds to step S108, whereas when the process is not completed, the process returns to step S102.
[0054]
In step S108, based on the free space after the previous frame processing of the buffer memory (not shown) connected to the subsequent stage, the information bit amount of the vertex value data of the block of 8 × 8 pixels, the histogram 1, etc., the block for 8 × 8 pixels is used. Calculate the threshold.
[0055]
Next, the residual table 1 is compared with the threshold calculated in step S108 (step S109), and it is determined whether the residual is within the threshold (step S110). If the residual is within the threshold, the process skips to step S118 described later.
[0056]
When the residual is not within the threshold value, in order to process a block of 4 × 4 pixels, in step S111, 4 is set as the number k of blocks when the block of 8 × 8 pixels is divided into blocks of 4 × 4 pixels. Set.
[0057]
Next, in step S112, the residuals of the four vertices of the 4 × 4 pixel block stored in the residual table 1 are quantized. For example, when the image is divided into blocks of 4 × 4 pixels as shown in FIG. 14C, A0′-1, A4-1, E0-1, and E4-1 become residuals of the target vertices. Next, in step S113, the residual of the vertex quantized in step S112 is inversely quantized to calculate a new residual. For example, new residuals (new residuals) of the residuals A0'-1, A4-1, E0-1 and E4-1 are A0 "-1, A4'-1, E0'-1 and E4 '. -1 (FIG. 14C).
[0058]
Next, in step S114, a new residual in a block of 4 × 4 pixels is calculated from the new residual of the four vertices calculated in step S113 and the bilinear surface, and the residual of each pixel is calculated. Find the residual. These residuals are represented by A0 ″ -2, A1-2,. . . D3-2 (FIG. 14D).
[0059]
Next, in step S115, the maximum value (absolute value) of the residual between the new residual and the residual of each pixel is detected, stored in the second residual table (residual table 2), and , In the second histogram (histogram 2).
[0060]
Next, 1 is subtracted from the number k of blocks (step S116), and thereafter, it is determined whether or not the processing on the block of 4 × 4 pixels has been completed (step S117). If the process has been completed, the process proceeds to step S118, whereas if not, the process returns to step S112.
[0061]
If the processing for the block of 4 × 4 pixels has been completed, or if it is determined in step S110 that the threshold value is larger than the residual, whether the processing for all blocks (residual table 1) has been completed in step S118 Determine whether or not.
[0062]
In step S119, a threshold value for the block of 4 × 4 pixels is calculated from the free space after the previous frame processing in the buffer memory, the information bit amount of the residual value data of the block of 4 × 4 pixels, the histogram 2 and the like. .
[0063]
Next, the residual table 2 is compared with the threshold calculated in step S119 (step S120), and it is determined whether the residual is within the threshold (step S121). If the residual is within the threshold, the process skips to step S129 described later.
[0064]
When the residual is not within the threshold value, in order to process a block of 2 × 2 pixels, in step S122, 4 is set as the number m of blocks when the block of 4 × 4 pixels is divided into blocks of 2 × 2 pixels. Set.
[0065]
Next, in step S123, the residuals of the four vertices of the 2 × 2 pixel block stored in the residual table 2 are quantized. For example, when the image is divided into 2 × 2 pixel blocks as shown in FIG. 14E, A0 ″ −2, A2-2, C0−2, and C2-2 become residuals of the target vertices. Next, in step S124, the vertex quantized in step S123 is inversely quantized to calculate a new vertex. For example, new residuals (new residuals) of the residuals A0 ″ -2, A2-2, C0-2, and C2-2 are A0 ′ ″-2, A2′-2, C0′-2 and C2′-2 (FIG. 14E).
[0066]
Next, in step S125, a new residual in a block of 2 × 2 pixels is calculated from the residual of the four vertices calculated in step S124 and the bilinear surface, and the new residual and the residual of each pixel are calculated. Find the residual with the difference. These residuals are A0 '''-3,A2'-3,C0'-3, and C2'-3 (FIG. 14F).
[0067]
Next, in step S126, the maximum value (absolute value) of the difference between the new residual and the residual of each pixel is detected and stored in a third residual table (residual table 3). No. 3 (histogram 3).
[0068]
Next, 1 is subtracted from the number m of blocks (step S127), and thereafter, it is determined whether or not the processing for the block of 2 × 2 pixels has been completed (step S128). If the process has been completed, the process proceeds to step S129. If the process has not been completed, the process returns to step S123.
[0069]
When the processing for the block of 2 × 2 pixels is completed or when it is determined in step S121 that the threshold value is larger than the residual, in step S129, all the blocks of 4 × 4 pixels (residual table 2) are processed. It is determined whether or not the process has been completed.
[0070]
In step S130, a threshold value for a 2 × 2 pixel block is calculated from the free space of the buffer memory after the previous frame processing, the information bit amount of the residual value data of the 2 × 2 pixel block, the histogram 3, and the like. .
[0071]
Next, the residual table 3 is compared with the threshold calculated in step S130 (step S131), and it is determined whether the residual is within the threshold (step S132). If the residual is within the threshold, the process skips to step S134 described later.
[0072]
If the residual is not within the threshold, the residual of four pixels is quantized (step S133), and thereafter, it is determined whether or not the processing for all the blocks of 2 × 2 pixels has been completed (step S134). When the process is completed, the control routine is terminated. When the process is not completed, the process returns to step S131.
[0073]
Note that the image decoding process on the receiving side of the image monitoring system is performed by performing an operation reverse to the flowcharts of FIGS.
[0074]
FIG. 15 is a block diagram of an embodiment of an image encoding device according to the present invention. This image coding apparatus is applied, for example, to the transmission side of an image monitoring system, and includes an input unit 1, an 8 × 8 block input circuit 2a, a 4 × 4 block input circuit 2b, and a 2 × 2 block input circuit 2c. , Prediction error quantizers 3a, 3b, 3c, prediction error dequantizer and new vertex calculation circuits 4a, 4b, 4c, new pixel calculation circuits 5a, 5b, 5c, and error calculation circuits 6a, 6b, 6c. And error comparing circuits 7a, 7b, 7c, a quantizer 8, a controller 9, and a switch 10.
[0075]
The operation of the present embodiment will be described. A digitized image signal from a surveillance camera (not shown) is input to the input unit 1, and the 8 × 8 block input circuit 2a divides the image into blocks of 8 × 8 pixels, and serves as a vertex of each block. A pixel (hereinafter, simply referred to as a vertex) is detected.
[0076]
The prediction error quantizer 3a calculates and quantizes the prediction error of the detected vertex. The quantization by the prediction error quantizer 3a reduces the number of data bits. The prediction error inverse quantizer and new vertex calculation circuit 4a inversely quantizes the error quantized by the prediction error quantizer 3a to obtain a new prediction error in order to correct the quantization error. A new vertex is calculated by adding or subtracting the prediction error to or from the vertex detected by the input circuit 2a.
[0077]
The new pixel calculation circuit 5a calculates a new pixel in the block using the new vertex calculated by the prediction error dequantizer and the new vertex calculation circuit 4a and the bilinear surface expressed by [Equation 1].
[0078]
The error calculation circuit 6a compares the new pixel calculated by the new pixel calculation circuit 5a with 64 pixels in the block to calculate an error. Each of the error comparison circuits 7a compares the error calculated by the error calculation circuit 6a with a threshold value. When the error is within the threshold value, the controller 9 is controlled so that the prediction error quantized by the prediction error quantizer 3a is output through the switch 10. On the other hand, when the error is outside the threshold, the block of 8 × 8 pixels is processed by the 4 × 4 block input circuit 2b and thereafter.
[0079]
The 4 × 4 block input circuit 2b further divides the block of 8 × 8 pixels into halves, that is, divides the block into 4 × 4 pixels having 4 pixels in the horizontal direction and 4 pixels in the vertical direction. A pixel serving as a vertex (hereinafter, simply referred to as a vertex) is detected.
[0080]
The prediction error quantizer 3b obtains and quantizes the prediction error of the vertex detected by the 4 × 4 block input circuit 2b excluding the vertex quantized by the prediction error quantizer 3a. The quantization by the prediction error quantizer 3b reduces the number of data bits. The prediction error inverse quantizer and new vertex calculation circuit 4b inversely quantizes the error quantized by the prediction error quantizer 3b to newly obtain a prediction error in order to correct the quantization error. A new vertex is calculated by adding or subtracting the prediction error to or from the vertex detected by the input circuit 2b.
[0081]
The new pixel calculation circuit 5b calculates a new pixel in the block using the new vertex calculated by the prediction error dequantizer and the new vertex calculation circuit 4b and the bilinear surface expressed by [Equation 1]. Note that the output of the prediction error inverse quantizer and the new vertex calculation circuit 4a is used for the vertex overlapping with the block of 8 × 8 pixels.
[0082]
The error calculation circuit 6b calculates an error by comparing the new pixel calculated by the new pixel calculation circuit 5b or the output of the prediction error dequantizer and the new vertex calculation circuit 4a with 16 pixels in the block. . Each of the error comparison circuits 7b compares the error calculated by the error calculation circuit 6b with a threshold value. When the error is within the threshold value, the controller 9 is controlled so that the prediction error quantized by the prediction error quantizer 3b is output through the switch 10. On the other hand, if the error is outside the threshold, the block of 4 × 4 pixels is processed by the 2 × 2 block input circuit 2c and thereafter.
The processing from the 4 × 4 block input circuit 2b to the error comparison circuit 7b is performed on four blocks.
[0083]
The 2 × 2 block input circuit 2c further divides the block of 4 × 4 pixels into halves, that is, divides the block into 2 × 2 pixels having 2 pixels in the horizontal direction and 2 pixels in the vertical direction. A pixel serving as a vertex (hereinafter, simply referred to as a vertex) is detected.
[0084]
The prediction error quantizer 3c calculates and quantizes the prediction error of the vertex detected by the 2 × 2 block input circuit 2c excluding the vertex quantized by the prediction error quantizer 3b. The number of data bits is reduced by the quantization by the prediction error quantizer 3c. The prediction error inverse quantizer and new vertex calculation circuit 4c inversely quantizes the error quantized by the prediction error quantizer 3c to newly obtain a prediction error in order to correct the quantization error. A new vertex is calculated by adding or subtracting the prediction error to or from the vertex detected by the input circuit 2c.
[0085]
The new pixel calculation circuit 5c calculates a new pixel in the block using the new vertex calculated by the prediction error dequantizer and the new vertex calculation circuit 4c and the bilinear surface expressed by [Equation 1]. For the vertices that overlap the block of 4 × 4 pixels, the outputs of the prediction error inverse quantizer and the new vertex calculation circuit 4b are used.
[0086]
The error calculating circuit 6c calculates an error by comparing the new pixel calculated by the new pixel calculating circuit 5c or the output of the prediction error dequantizer and the new vertex calculating circuit 4b with the four pixels in the block. . Each of the error comparison circuits 7c compares the error calculated by the error calculation circuit 6c with a threshold value. When the error is within the threshold value, the controller 9 is controlled so that the prediction error quantized by the prediction error quantizer 3c is output through the switch 10. On the other hand, if the error is outside the threshold, the quantizer 8 quantizes the error calculated by the error calculation circuit 6c.
[0087]
The signal output from the switch 10 is input on the receiving side to an image decoding device configured by elements similar to the elements illustrated in FIG. 15, and performs an operation reverse to the operation of the image encoding device illustrated in FIG. 15. By doing so, it is decoded and output to, for example, a monitor.
[0088]
According to the above embodiment, the surveillance camera uses a bilinear surface of block processing of 8 × 8 pixels for a flat or gentle slope such as a road or a wall, and uses a bi-linear surface for undulating parts. By performing block processing of 2 × 2 pixels, a suitable image monitoring system can be configured. Furthermore, when the flat or gentle slope is relatively large, a transmission system capable of further increasing the compression ratio by using a block of 16 × 16 pixels larger than a block of 8 × 8 pixels is to be constructed. Can be.
[0089]
The present invention is not limited to the above embodiment, and many modifications and variations are possible.
For example, the image coding method and device according to the present invention can be used on the transmission side of any transmission system, and the corresponding image decoding method and device can be used on the reception side of the corresponding system. .
[0090]
In addition, the processing was performed in the order of the block of 8 × 8 pixels, the block of 4 × 4 pixels, and the block of 2 × 2 pixels. ⁿ × 2 ⁿ It can be applied to a block of pixels (n: natural number). For example, processing is performed in the order of a block of 16 × 16 pixels, a block of 8 × 8 pixels, a block of 4 × 4 pixels, and a block of 2 × 2 pixels. You can also.
[0091]
In the above embodiment, the process of quantizing the vertices of the block of 8 × 8 pixels, the block of 4 × 4 pixels, and the block of 2 × 2 pixels and converting them into variable-length bits was performed. By performing transmission using fixed-length bits, the amount of calculation can be reduced.
[0092]
Further, the case of the intra-frame processing has been described, but the inter-frame processing may be combined. That is, when processing a block of an arbitrary size, an inter-frame process can be easily added by inserting a process for comparing with a previous frame.
[Brief description of the drawings]
FIG. 1 is a diagram showing vertices of a divided block.
FIG. 2 is a flowchart of an embodiment of an image encoding method according to the present invention.
FIG. 3 is a part of a flowchart of a subroutine of step S1 in FIG. 2;
FIG. 4 is a part of a flowchart of a subroutine of step S1 in FIG. 2;
FIG. 5 is a diagram for explaining calculation of a new pixel.
6 is a diagram showing histograms 1, 2, and 3 created in the flowchart of FIG.
FIG. 7 is a part of a flowchart of a subroutine of step S3 in FIG. 2;
FIG. 8 is a part of a flowchart of a subroutine of step S3 in FIG. 2;
FIG. 9 is a diagram for explaining calculation of a new vertex.
FIG. 10 is a part of a flowchart of another embodiment of the image encoding method according to the present invention.
FIG. 11 is a part of a flowchart of another embodiment of the image encoding method according to the present invention.
FIG. 12 is a part of a flowchart of another embodiment of the image encoding method according to the present invention.
FIG. 13 is a part of a flowchart of another embodiment of the image encoding method according to the present invention.
FIG. 14 is a diagram for explaining a residual of a vertex of a block.
FIG. 15 is a block diagram of an embodiment of an image encoding device according to the present invention.
[Explanation of symbols]
1 Input section
2a 8 × 8 block input circuit
2b 4 × 4 block input circuit
2c 2 × 2 block input circuit
3a, 3b, 3c prediction error quantizer
4a, 4b, 4c prediction error inverse quantizer and new vertex calculation circuit
5a, 5b, 5c New pixel calculation circuit
6a, 6b, 6c Error calculation circuit
7a, 7b, 7c error comparison circuit
8 Quantizer
9 Controller
10 Switch

Claims (3)

An image encoding method for encoding and transmitting an image signal, comprising:
The entire screen is divided into predetermined pixel units,
Using the vertices and bilinear surfaces of the pixels in the block, calculate a new pixel value,
Calculate the difference between the pixel value in the block and the new pixel value,
Based on whether or not these differences are within a predetermined range, determine whether to encode the block or to further divide the block,
An image encoding method, comprising: performing an encoding process on an appropriate pixel-unit block according to a result of the determination. 2. The image encoding method according to claim 1, wherein a histogram in which a maximum value of the absolute value of the difference is obtained for each block is created, and the predetermined value is calculated from the histogram. An image encoding device for encoding and transmitting an image signal,
Block dividing means for dividing the entire screen into blocks of a predetermined pixel unit,
New pixel value calculation means for calculating a new pixel value using the vertices and bilinear surfaces of the pixels in the block,
A pixel value in the block, and a difference calculation unit for calculating a difference between the new pixel value,
A determination unit configured to determine whether to encode the block or to further divide the block, based on whether these differences are within a predetermined range;
An image encoding device comprising: an encoding unit that performs an encoding process on a block in an appropriate pixel unit according to a result of the determination.

Images