JP2000138937A - Image compression coding method and device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To enhance the compression efficiency and to suppress deterioration in the image quality. SOLUTION: The image compression device is an MPEG image compression device that respectively selects in-frame compression or inter-frame compression for each macro block, the compression device has a motion detection section 11 that obtains a mean square error for each of block divisions that are 1/4 divisions of a macro block and controller 13 that sets priority for selection of the in-frame compression or the inter-frame compression and selects the in-frame compression as a macro block type when a maximum value of the means square errors of each block is sufficiently small.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to a so-called M
PEG (Moving Picture image coding Experts Group)
The present invention relates to an image compression encoding method and apparatus for performing motion detection such as the above.

[0002]

2. Description of the Related Art In the case of performing inter-frame compression using motion detection such as MPEG, an image in one frame as shown in FIG. This is appropriately referred to as a macroblock MB), and motion detection from the reference image to the current image is performed in units of the macroblock MB as shown in FIG. After performing the motion detection, a macroblock obtained by compressing the residual information and the motion vector from the reference image to the current image between frames is called an inter macroblock. A macroblock that compresses only the information of the current image is called an intra macroblock.

FIG. 5 shows the configuration of a general MPEG image compression apparatus.

In FIG. 5, data of a current image is input to a terminal 100 and sent to a motion detecting unit 101.

[0005] The motion detection unit 101 performs motion detection on a macroblock basis using the data of the current image and the data of the previous image (reference image) stored in the memory 102. It outputs a mean square prediction error and a mean square deviation described later.

The controller 103 is composed of a DSP (Digital Signal Processor), and uses the information of the mean square prediction error and the mean square deviation from the motion detection 101 to determine the intra-frame compression or the inter-frame compression for each macro block. That is, it is determined whether it is an intra macro block or an inter macro block. Each macro block is intra-frame compression or inter-frame compression (intra macro block or inter macro block)
Is determined as the macroblock type, and the information of the macroblock type is output from the controller 103 to the motion detection unit 101.

[0007] Thereafter, from the motion detecting section 101, image data for an intra macroblock or a motion prediction residual component for an inter macroblock and a motion vector are output according to the macroblock type from the controller 103. The signal is sent to a DCT (discrete cosine transform) and quantization unit 104.

[0008] The DCT and quantization unit 104 performs DCT on the image data in the case of an intra macroblock and the motion prediction residual component in the case of an inter macroblock in units of macroblocks, and furthermore, the quantization scale determined in units of macroblocks. , And sends the quantized DCT coefficient to the variable-length coding unit 105 together with the motion vector.

The variable length coding unit 105 performs variable length coding on the data supplied from the DCT and quantization unit 104,
The obtained compressed data is output from the terminal 106.

As described above, in MPEG and the like, since an image is formed by mixing intra / inter macroblocks in the same frame, it is determined whether each macroblock uses an intra macroblock or an inter macroblock. It must be determined at the time of image compression.

In this case, it is necessary to select a macroblock type having good coding efficiency and little deterioration in image quality to perform coding.

In the case of the conventional technique shown in FIG. 5, for example, the following method is used when determining an intra / inter macro block.

The process of determining an intra / inter macro block in the configuration of FIG. 5 will be described with reference to FIGS. 6 and 7.

First, as shown in FIG. 6 and Equations (1) and (2), the motion detection unit 101 sets each luminance value Y (x, y) (16 (16) pixels of the macroblock of the current image to 16 × 16 pixels. Y (0,
0)) to the average value avgY of Y (15,15)), and the average value avgY
The difference between Y and the luminance value Y (x, y) of each pixel is squared and added to obtain an average. Hereinafter, this is called the mean square deviation, and the symbol VA
Expressed as ROR.

[0015]

(Equation 1)

Further, as shown in FIG. 6 and equation (3), the motion detecting section 101 sets, in the macroblock of the reference picture corresponding to the current picture at the time of motion prediction, the macroblock between the current picture and the reference picture. The square of the difference between each of the luminance values Y (0,0) to Y (15,15) and Y ′ (0,0) to Y ′ (15,15) of each pixel having the same coordinates is obtained, and this is 16 × 16 pixels are added and the average is obtained. Hereinafter, this is referred to as the mean square prediction error.
Expressed as VAR.

[0017]

(Equation 2)

Next, the mean square deviation VAROR and mean square prediction error VAR obtained by the motion detecting section 101 are sent to the controller 103 as step S102 in FIG.

Next, in the controller 103, at step S103, the mean square deviation VAROR and the mean 2
The power prediction error VAR is compared. If VAR ≦ VAROR, an inter macroblock is selected in step S105, and the macroblock type is output to the motion detection unit 101.
On the other hand, if VAROR <VAR, an intra macroblock is selected in step 104, and the macroblock type is output to the motion detection unit 101.

Thereafter, when the macroblock types for all the macroblocks have been determined in step S106, the process proceeds to the next frame in step S107. When the types of all the macroblocks have not been determined, the process proceeds to step S107. The processing after S102 is performed.

As described above, intra-frame compression intra or intra-frame compression is performed based on each macroblock type.

[0022]

As described above, when inter-frame compression is performed using motion detection as described above and when only intra-frame compression is performed, generally, inter-frame compression is more efficient than data compression. It is known that the amount is small and the compression efficiency is good. However, as shown in the flowchart of the conventional macroblock type determination processing shown in FIG. 7, the intra / inter macroblock type discrimination is performed simply by simply comparing the mean square deviation VAROR and the mean square prediction error VAR. Therefore, the number of intra macroblocks selected is large, and the compression efficiency is not good.

On the other hand, it is also known that, for example, on a flat screen, deterioration of image quality is easily noticeable to human eyes. However, in the case of the conventional example described above, if the inside of the macroblock is flat, the difference between the mean square deviation VAROR and the mean square prediction error VAR will occur even if the motion detection fails and an incorrect motion vector is obtained. There is a case where an inter macroblock having a wrong motion vector is difficult to appear and is selected. In this case, since the screen is flat, the deterioration of the image quality tends to be conspicuous. In particular, as described above, if the inter-macroblock is simply selected with a view to increasing the compression efficiency, this problem becomes conspicuous.

Accordingly, the present invention has been made in view of such a situation, and provides an image compression encoding method and apparatus capable of increasing compression efficiency and suppressing deterioration of image quality. With the goal.

[0025]

According to the present invention, there is provided an image compression / encoding method for performing compression by selecting either intra-image compression or inter-image compression for each encoding unit including a plurality of pixels. Priority is given to the selection of intra-image compression and inter-image compression, a predetermined parameter is obtained for each block obtained by dividing the coding unit into a plurality, and a predetermined parameter corresponding to the predetermined parameter and visual characteristics of each block is obtained. The above-described problem is solved by changing the priority based on the comparison with the value and selecting the intra-image compression or the inter-image compression based on the priority.

An image compression encoding apparatus according to the present invention is an image compression encoding apparatus which performs compression by selecting either intra-image compression or inter-image compression for each encoding unit composed of a plurality of pixels. Priority setting means for setting priorities used for selection of intra-compression and inter-picture compression; parameter generation means for obtaining a predetermined parameter for each block obtained by dividing the coding unit into a plurality of blocks; A comparing unit for comparing with a predetermined value corresponding thereto, a priority control unit for changing a priority based on the comparison result, and a selecting unit for selecting intra-image compression or inter-image compression based on the priority. Solves the above-described problem.

[0027]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Preferred embodiments of the present invention will be described with reference to the drawings.

FIG. 1 shows the configuration of an MPEG image compression apparatus as an embodiment to which the image compression encoding method and apparatus of the present invention is applied.

In FIG. 1, data of the current image is input to a terminal 10 and sent to a motion detecting section 11.

The motion detecting section 11 performs motion detection on a macroblock basis using the data of the current image and the data of the previous image (reference image) stored in the memory 12, and obtains the equations (1) to (4). The mean square deviation VAROR and the mean square prediction error VAR are obtained by the equation (3), and the mean square deviation VAROR and the mean square prediction error VAR are output to the controller 13.

Further, in the motion detecting section 11 of the present embodiment, as shown in FIG. 2 and equations (4) and (5),
4 8 × 8 blocks bl of 16-pixel macroblock
divided into ock1 ~ block4, each block block1 ~ bloc
At k4, each luminance value of 8 × 8 pixels Y (x, y) (Y (0,0)) to
Y (7,7)) is calculated, and the difference between the average value avgy and the luminance value Y (x, y) of each pixel is squared and the average is calculated. The mean square deviation of each of the blocks block1 to block4 represents the flatness of each of the blocks block1 to block4, and is hereinafter represented by the symbols ACTIVITY (1) to ACTIVITY (4). Equation (4)
And equation (5) are, for example, the mean square deviation AC of block block1.
This is an expression for calculating TIVITY (1).

[0032]

(Equation 3)

The motion detector 11 divides the macroblock into four blocks 1 to block, as described above.
The mean square deviations ACTIVITY (1) to ACTIVITY (4) of k4 are also output to the controller 13.

The controller 13 is composed of a DSP (Digital Signal Processor), and has a mean square deviation VAROR and a mean square prediction error VAR from the motion detection 11, and an average 2 of each block block1 to block4 obtained by dividing the macroblock into four.
Using the information of the power deviations ACTIVITY (1) to ACTIVITY (4), and further using, for example, T1, T2, P1, and P2 which are fixed constants described later set from the terminal 17, as described later, each macro Judgment of intra-frame compression or inter-frame compression is performed for each block, that is, determination of intra macroblock or inter macroblock. It should be noted that the fixed values T1 and T2 are, as described later,
It is a constant that is compared with the maximum value of the mean square deviations ACTIVITY (1) to ACTIVITY (4) of k1 to block4, is stored as internal data of the controller 13, and has a relationship of T1> T2. The fixed values P1 and P2 are constants used to give priority (weight) to the macroblock type, and are stored as internal data of the controller 13. These fixed values T1, T2, P1,
P2 is determined in advance in consideration of visual characteristics.

The controller 13 determines whether each macroblock is intra-frame compression or inter-frame compression (intra macroblock or inter-macroblock) as a macroblock type, and the information of the macroblock type is determined by the controller 1.
3 to the motion detector 11.

Thereafter, from the motion detecting section 11, image data for an intra macro block or a motion prediction residual component for an inter macro block and a motion vector are output according to the macro block type from the controller 13. The signal is sent to a DCT (discrete cosine transform) and quantization unit 14.

The DCT and quantization section 14 performs DCT on the image data in the case of an intra macroblock or the motion prediction residual component in the case of an inter macroblock on a macroblock basis, and furthermore, a quantization scale determined on a macroblock basis. , And sends the quantized DCT coefficient to the variable-length encoding unit 15 together with the motion vector.

The variable length coding unit 15 performs variable length coding on the data supplied from the DCT and quantization unit 14 and outputs the obtained compressed data from a terminal 16.

As described above, when inter-frame compression is performed using motion detection and when only intra-frame compression is performed, inter-frame compression generally has a smaller data amount and higher compression efficiency. It is known. However, as in the above-mentioned conventional example, simply the mean square deviation VARO
When intra / inter macroblock type discrimination is performed only by comparing R and the mean square prediction error VAR, the number of intra macroblocks selected is large and the compression efficiency is not good.

From the above, in the present embodiment,
In order to make it easier to select inter macroblocks regularly and to improve the efficiency of image compression, make it easier to select inter macroblocks when comparing the mean square deviation VAROR and the mean square prediction error VAR. A fixed value P1 is provided, and when the mean square prediction error VAR is smaller than the fixed value P1, an inter macroblock is selected.

As described above, for example, on a flat screen, the deterioration of the image quality is easily noticeable to human eyes. For example, when the inside of a macro block is flat, the motion detection is not successful and an incorrect motion vector is obtained. Even if you bring the difference between the mean square deviation VAROR and the mean square prediction error VAR,
An inter macroblock having an incorrect motion vector may be selected. In this case, since the screen is flat, deterioration in image quality is conspicuous as deterioration. In particular, if priority is simply given to an inter macro block in consideration of increasing compression efficiency, this problem becomes conspicuous.

From the above, in the present embodiment,
If the motion detection is not successful on a flat screen, it indicates the degree of flatness within the macroblock so that intra macroblocks can be easily selected and image quality degradation in visual characteristics can be prevented. Average value of each block block1 to block4 obtained by dividing a macroblock into four as a value
The squared deviations ACTIVITY (1) to ACTIVITY (4) are obtained, and the maximum mean squared deviation ACTIVITY is used. The maximum mean squared deviation ACTIVITY is smaller than a certain fixed value T1. If the squared deviation ACTIVITY is smaller than a certain fixed value T2, a value obtained by adding a certain fixed value P2 and a mean square prediction error VAR (VAR + P2) and a mean squared deviation VAROR so that an intra macroblock is easily selected. Comparisons are made between them.

In the present embodiment, the above-described macro block selection processing is realized by a flowchart as shown in FIG.

In FIG. 3, first, in the motion detecting section 11,
In step S1, the above equations (1) to (5) are calculated, and the mean square deviation VAROR in the macro block is calculated.
And the mean square prediction error VAR and the mean square deviation AC of each of the blocks block1 to block4 obtained by dividing the macroblock into four.
Find TIVITY (1)-ACTIVITY (4). These mean square deviations
VAROR and mean square prediction error VAR and four mean square deviations AC
The data of TIVITY (1) to ACTIVITY (4)
Is passed to the register.

The controller 13 has a step S
2, fixed values T1, T2, P1, and P2 are input in advance and stored as internal data of the controller 13.

The controller 13 performs, at step S3, the four mean square deviations ACTIVITY stored in the register.
The largest one is selected from (1) to ACTIVITY (4), and the largest mean square deviation ACTIVITY is stored in a register as a variable ACT.

Next, the controller 13 proceeds to step S4
The variable ACT is compared with the fixed value T1, and if ACT <T1, the process proceeds to step S6. If ACT ≧ T1, the process proceeds to step S5.

When the process proceeds to step S5, the controller 13 compares the mean square prediction error VAR with the fixed value P1. If the average square prediction error VAR is smaller than the fixed value P1 in the comparison in step S5, the process proceeds to step S11, and the macroblock type is forcibly determined to be an inter macroblock. Accordingly, when the flatness is low, that is, when the maximum mean square deviation ACTIVITY is high, the ratio of selecting the inter macroblock as the macroblock type increases. On the other hand, step S5
If the mean square prediction error VAR is larger than the fixed value P1, the process proceeds to step S6.

When the process proceeds to step S6, the controller 13 compares the variable ACT with the fixed value T2. In this comparison, when ACT ≧ T2 (or ACT> T2),
Proceeding to step S7, the magnitude of the mean square deviation VAROR and the magnitude of the mean square prediction error VAR are compared without assigning a priority to either the inter or the intra. Thus, if the maximum mean square deviation ACTIVITY is low to some extent,
Intra macroblock types are equally selected.
That is, in step S7, if VAR ≦ VAROR, an inter macroblock is selected in step S11. On the other hand, if VAR> VAROR, step S1 is performed.
At 2, an intra macroblock is selected.

On the other hand, in step S6, ACT <T2
Is satisfied, the process of the controller 13 proceeds to step S8, and the mean square prediction error VAR is set so that the intra macroblock is prioritized as the macroblock type.
(VAR + P2) obtained by adding the fixed value P2 to the mean square deviation VAROR
Compare large and small. This gives the maximum mean squared deviation AC
If the TIVITY is sufficiently low, the intra macroblock has priority. If the result of the magnitude comparison in step S8 is that the mean square deviation VAROR is smaller than the prioritized mean square prediction error (VAR + P2), the macroblock type is set to an intra macroblock in step S9. Otherwise, in step S10, the macro block type is set to an inter macro block.

The macroblock type set by the controller 13 as described above is passed to the motion detector 11 in step S13. Thereafter, in step S14, when the macroblock types for all the macroblocks have been determined, the process proceeds to the next frame in step S15, and when the type of all the macroblocks has not been determined, the processes after step S1 are performed. I do. Thereby, in the configuration after the motion detection unit 11, the controller 1
Intra-frame or inter-frame compression is performed based on the macroblock type determined in step 3.

As described above, according to the embodiment of the present invention, the overall image quality can be improved by preferentially using the pixel blocks of the inter-frame compression having high compression efficiency in the steady state. In addition, in a flat pixel block in which the image quality is conspicuous in visual characteristics, the priority is given to intra-frame compression. Can be prevented, and both efficient compression and high image quality can be achieved.

[0053]

As is apparent from the above description, in the present invention, priority is given to the selection between intra-image compression and inter-image compression, and a predetermined coding unit is obtained for each block obtained by dividing the coding unit into a plurality of blocks. By changing the priority based on the comparison between the parameter and a predetermined value corresponding to the visual characteristic, and selecting the intra-image compression or the inter-image compression based on the priority, the compression efficiency is increased and the deterioration of the image quality is reduced. Can also be suppressed.

In other words, according to the present invention, in the steady state, the coding quality of the inter-picture compression having high compression efficiency is preferentially used to improve the overall picture quality, and the visual quality can reduce the picture quality. By giving priority to intra-image compression for conspicuous flat coding units, it is possible to prevent image quality degradation even when motion detection accuracy is poor, compared to the case of simple inter-image compression priority selection. It achieves both effective compression and high image quality.

[Brief description of the drawings]

FIG. 1 is a block circuit diagram showing a schematic configuration of an MPEG image compression apparatus as an embodiment to which an image compression encoding method and apparatus of the present invention is applied.

FIG. 2 is a diagram used to explain each block obtained by dividing a macroblock into four and the mean square deviation of each block;

FIG. 3 is a flowchart illustrating a flow of a macroblock type selection process in the MPEG image compression apparatus according to the present embodiment.

FIG. 4 is a diagram used for describing macroblocks and inter-frame prediction.

FIG. 5 is a block circuit diagram showing a schematic configuration of a conventional MPEG image compression apparatus.

FIG. 6 is a diagram used to describe a mean square deviation and a mean square prediction error.

FIG. 7 is a flowchart showing the flow of a macroblock type selection process in a conventional MPEG image compression apparatus.

[Explanation of symbols]

11 motion detection unit, 12 memory, 13 controller, 14 DCT and quantization unit, 15 variable length coding unit

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5C059 KK02 MA00 MA05 MA23 MC11 MC38 ME01 NN01 NN28 PP05 PP06 TA18 TA22 TB07 TB08 TC02 TD04 TD11 UA02 5J064 AA02 BA13 BB03 BC02 BD03

Claims (6)

[Claims] 1. An image compression coding method for performing compression by selecting intra-image compression or inter-image compression for each encoding unit including a plurality of pixels, wherein the intra-image compression and the inter-image compression are prioritized. A predetermined parameter is obtained for each block obtained by dividing the coding unit into a plurality of parts, and the priority is changed based on a comparison between the predetermined parameter for each block and a predetermined value according to visual characteristics. An image compression / encoding method, wherein the intra-image compression or the inter-image compression is selected based on the priority. 2. The image compression encoding method according to claim 1, wherein an average square error of pixel values in each block is obtained as the predetermined parameter. 3. The maximum value of the mean square error of each block is:
3. The image compression encoding method according to claim 2, wherein when the value is smaller than a predetermined value corresponding to the visual characteristic, the intra-image compression is selected. 4. An image compression encoding apparatus for performing compression by selecting either intra-image compression or inter-image compression for each encoding unit including a plurality of pixels, wherein the intra-image compression and the inter-image compression are selected. Priority setting means for setting a priority; parameter generating means for obtaining a predetermined parameter for each block obtained by dividing the coding unit into a plurality of blocks; and comparing the predetermined parameter with a predetermined value according to visual characteristics. Comparison means for performing, priority control means for changing the priority based on the comparison result, and selection means for selecting the intra-image compression or the inter-image compression based on the priority. Image compression encoding device. 5. The image compression encoding apparatus according to claim 4, wherein said parameter generating means obtains a mean square error of pixel values in each block as said predetermined parameter. 6. The apparatus according to claim 5, wherein said selecting means selects said intra-image compression when the maximum value of the mean square error of each block is smaller than a predetermined value corresponding to said visual characteristics. Image compression encoding device.