JP2000165878A - Device and method for encoding - Google Patents

Abstract

PROBLEM TO BE SOLVED: To efficiently compress data while effectively avoiding the degradation of picture quality even concerning a high definition video signal by providing a coefficient correcting means for correcting a coefficient value for the unit of each fine block with an average value as a reference and a quantizing means, etc., for quantizing the coefficient data corrected by this means. SOLUTION: A coefficient processing circuit 23 calculates the average value of absolute values of coefficient values for the unit of a macro block concerning coefficient data D1 and outputs coefficient data D2. An inverse discrete cosine transformation(IDCT) circuit 24 performs IDCT processing to the coefficient data D2 outputted from the coefficient processing circuit 23 and outputs image data to a subtracter 3. Then, a discrete cosine transformation(DCT) circuit 5 performs DCT processing to the image data inputted to the subtracter 3, and these data are quantized by a quantizing circuit 6. Thus, since the average value is calculated and the coefficient data are corrected with the average value as the reference, even concerning the high definition video signal, the degradation of picture quality can be effectively avoided.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an encoding apparatus and an encoding method, for example, an MPEG (Moving Picture Exposure).
(erts Group) can be applied to the case of encoding a video signal. The invention uses DCT
Calculates the average value of the absolute value of coefficient data for each small block obtained by subdividing the block, and corrects the coefficient data based on this average value to effectively reduce the deterioration of image quality even for high-definition video signals. Avoid data compression efficiently.

[0002]

2. Description of the Related Art Conventionally, in an encoding apparatus using the MPEG format, a video signal is compressed at a desired bit rate by setting a quantization scale value in consideration of human visual characteristics.

FIG. 7 is a block diagram showing this type of coding apparatus. In the encoding device 1, the preprocessing circuit 2 rearranges the pictures of the sequentially input image data DV1 in the encoding order. Further, the preprocessing circuit 2
A macroblock is set for each picture, and image data DV1 sequentially input in the order of raster scan is output in the order of block scan.

At this time, the pre-processing circuit 2 is, for example, 4: 2:
When data processing is performed on the image data DV1 in the 0 format, the chrominance data is thinned out according to the format of the input image data DV1. Set so that DCT blocks correspond. Here, the DCT block is a processing unit of the discrete cosine transform processing, and is a block of 8 × 8 pixels in MPEG. As a result, the preprocessing circuit 2
One macroblock is composed of four blocks of luminance data of × 8 pixels and two corresponding color difference blocks Cb and Cr.

The subtractor 3 receives the image data output from the pre-processing circuit 2 and subtracts the image data output from the motion compensation circuit 4 to output the subtracted data. The output image data is output to the subsequent discrete cosine transform circuit (DCT) 5 without any processing, whereas the output data of the pre-processing circuit 2 and the output from the motion compensation circuit 4 are output for P and B pictures. The difference data from the predicted frame to be output is output to the discrete cosine transform circuit 5.

The discrete cosine transform circuit 5 performs a discrete cosine transform process on the image data input from the subtracter 3 in DCT block units, and outputs the resulting coefficient data to a quantization circuit 6. The quantization circuit 6 quantizes the coefficient data output from the discrete cosine transform circuit 5 according to the quantization scale value set by the rate control circuit 7 and outputs the result.

[0007] The variable length encoding circuit 8 is
Is subjected to variable-length encoding and output. At this time, the variable-length encoding circuit 8 performs variable-length encoding on the output data value of the quantization circuit 6 that has a statistically high frequency so as to have a short code length, thereby reducing the amount of generated data. I do.

The buffer 9 accumulates the output data of the variable length encoding circuit 8 and outputs it at a constant data transfer rate. At this time, the buffer 9 stores a motion vector detected and encoded by a motion vector detection circuit (not shown),
The output data of the variable length encoding circuit 8 is output in a predetermined format together with the quantization table and the picture type data in the quantization circuit 6. In the encoding device 1, the output data of the buffer 9 is supplied as a video stream to, for example, a multiplexer.

The inverse quantization circuit 12 decodes the input data of the quantization circuit 6 by subjecting the output data of the quantization circuit 6 to inverse quantization, and the inverse discrete cosine transform circuit (inverse DCT) 13 applies The input data of the discrete cosine transform circuit 5 is decoded by subjecting the output data of the quantization circuit 12 to inverse discrete cosine transform processing.

The adder circuit 14 stores the output data of the inverse discrete cosine transform circuit 13 in the frame memory 15 as it is for the I picture, so that the image data necessary for prediction of a P picture or the like following the frame memory 15 is stored. Hold. Addition circuit 14
Reproduces the image data input to the subtractor 3 by adding the output data of the inverse discrete cosine transform circuit 13 and the predicted frame output from the motion compensation circuit 4 in the P picture and the B picture,
The reproduced image data is stored in the frame memory 15.

The motion compensation circuit 4 reads image data from the frame memory 15 at a timing corresponding to the motion vector, and outputs the read image data to the subtracter 3 as image data of a predicted frame.

The rate control circuit 7 monitors an empty area of the buffer 9, and according to the empty area, for example, TM5 rate control (ISO / IEC JTC1 / SC29 / WG11 / NO400 Test Mode).
del5, PP54-57 (April, 1993)) controls the quantization scale value of the quantization circuit 6.

That is, the rate control circuit 7 includes a buffer 9
The amount of bits to be allocated to each frame is calculated from the empty area of, and the amount of bits to be allocated to each macroblock is calculated from the amount of bits allocated to each frame. The rate control circuit 7 temporarily sets a quantization scale value in macroblock units based on the bit amount allocated to each macroblock.

The rate control circuit 7 calculates a weighting factor activity in consideration of visual characteristics based on a frame prediction residual or the like detected when detecting a motion vector.
The quantization scale value temporarily set by vity is modulated to calculate the quantization scale value.

Thus, the rate control circuit 7 sets the temporarily set quantization scale value to Q [n], and sets the normalized coefficient of the weight coefficient activity in consideration of the visual characteristics to nact.
Assuming [n], a final quantization scale mQ [n] is calculated by the following equation.

[0016]

(Equation 1)

Thus, in the quantization circuit 6,
For example, with respect to the quantization matrix set so that the quantization step size increases as the frequency component increases as shown in FIG.
The weighting process is performed by Q [n], and the coefficient data output from the discrete cosine transform circuit 5 is sequentially quantized by the resulting quantization matrix.

[0018]

By the way, in a quantization matrix which is a reference of such a quantization process, a portion corresponding to higher-order coefficient data has a larger quantization step size. Therefore, when quantization is performed using such a quantization matrix, in a decoded image, the higher the frequency components, the more the image deteriorates.

In a video signal supplied by ordinary television broadcasting or the like, such deterioration of high-frequency components having low visual sensitivity is not a problem because it is hard to perceive.

However, in the case of encoding a high-definition image, the generated code amount increases and the quantization scale value becomes large, so that the deterioration of the high frequency components becomes remarkable as compared with the original image. , The deterioration of the high frequency component is perceived.

On the other hand, recently, in this type of encoding processing, for example, HDTV (High Definition TeleVisio)
n), it is required to encode a higher-definition video signal as in the case of transmission, and the processing target itself, such as the imaging result of a digital camera, has been improved by improving the performance of the image input device. High definition.

SUMMARY OF THE INVENTION The present invention has been made in view of the above points. An encoding apparatus and an encoding method capable of effectively avoiding deterioration of image quality and efficiently compressing even a high-definition video signal. It is intended to propose.

[0023]

According to the present invention, in order to solve the above-mentioned problems, the present invention is applied to an encoding device or an encoding method, and for coefficient data obtained by discrete cosine transform, each minute block obtained by dividing a matrix-like array is used. Then, the average value of the absolute values of the coefficient values is calculated, and the coefficient value is corrected for each minute block based on the average value.

Further, the average value of the absolute values of the coefficient data is calculated in units of blocks for the coefficient data obtained by the discrete cosine transform processing by applying the present invention to the encoding apparatus or the encoding method. The average value of the absolute values of the coefficient data is calculated, the average value for each block is corrected with the average value in the immediately preceding frame, and the quantization step size in quantization is varied according to the correction result.

If the average value of the absolute values of the coefficient values is calculated for each of the small blocks obtained by dividing the matrix-like array in the coefficient data, the characteristic point in the discrete cosine transform processing target can be calculated based on the average value. It can be determined whether they belong. Thus, if the coefficient value is corrected in units of minute blocks based on the average value, the amount of information can be reduced by effectively avoiding the deterioration of feature points due to the coefficient value in units of blocks subjected to discrete cosine transform processing. be able to. As a result, even for a high-definition video signal, it is possible to effectively avoid image quality degradation and efficiently compress data.

If the average value in blocks is corrected by the average value in the immediately preceding frame and the quantization scale value in quantization is varied according to the correction result, the average value in blocks is large. In such a case, the quantization scale value can be reduced. Therefore, a large data amount can be assigned to a portion where the generated code amount is large, and the data can be efficiently compressed without effectively deteriorating the image quality.

[0027]

Embodiments of the present invention will be described below in detail with reference to the drawings.

(1) First Embodiment (1-1) Configuration of First Embodiment FIG. 1 is a block diagram showing an encoding apparatus according to a first embodiment of the present invention. In this encoding device 20,
The same components as those of the encoding device 1 described above with reference to FIG. 7 are denoted by the corresponding reference numerals, and redundant description will be omitted.

In the encoding device 20, a discrete cosine transform circuit 22, a coefficient processing circuit 23, and an inverse discrete cosine transform circuit 24 are arranged between the preprocessing circuit 2 and the subtractor 3. The information amount of the image data DV1 is reduced, and the rate control circuit 2
5 controls the quantization scale.

Here, the discrete cosine conversion circuit 2
Reference numeral 2 denotes a discrete cosine transform process on the image data DV1 input in the block scan order from the preprocessing circuit 2 in units of a DCT block, which is a block of 8 × 8 pixels, and outputs coefficient data D1 obtained as a result. I do.

The coefficient processing circuit 23 is constituted by, for example, a digital signal processor.
Is corrected to reduce the data amount of the image data DV1.
Is output.

Here, as shown in FIG. 2, the coefficient data D1 obtained by the discrete cosine transform process corresponds to a DCT block of 8 × 8 pixels, and sequentially represents frequency components in the horizontal and vertical directions in this DCT block. 8 × 8 coefficient values.

The coefficient processing circuit 23 divides the matrix-like arrangement of the 8 × 8 coefficient values equally into the high frequency side and the low frequency side in the horizontal direction and the vertical direction. Four blocks A to D are set by numerical values, and average values AVA to AVD of the absolute values of the coefficient values are calculated for each of the blocks A to D.

The coefficient processing circuit 23 calculates the average value AVA〜
A coefficient value setting process for setting the coefficient values of predetermined blocks A to D to a predetermined value based on the AVD is executed. Further, the coefficient processing circuit 23 performs a gain coefficient calculation process based on the average values AVA to AVD, and weights a coefficient value obtained by the coefficient value correction process with the gain coefficient.

In the calculation of the average value, the coefficient processing circuit 23 excludes the 0th-order horizontal and 0th-order coefficients indicating the level of the DC component from the AC coefficient value for the block A on the lowest frequency side. Calculate the average value from Also, in the coefficient value setting process, for the block A on the lowest frequency side, coefficients are set only for AC coefficient values other than the 0th-order horizontal and 0th-order coefficients indicating the level of the DC component. Prevents image quality degradation due to level changes.

That is, in the coefficient value setting process, the coefficient processing circuit 23 detects a block whose average value AVA to AVD is equal to or smaller than a predetermined value t, and sets all coefficient values to a value 0 for this block. Here, the predetermined value t
Is set according to the resolution required for the encoding result, and in this embodiment, is set to, for example, the value 5. As a result, the coefficient processing circuit 23 sets the coefficient to a value of 0 for a block in which the degree of coefficient generation is small.
At this time, for the block A on the lowest frequency side, even if this condition is satisfied, the process of setting the coefficient to the value 0 is excluded, thereby preventing the image quality deterioration due to the removal of the low frequency component. You.

Subsequently, the coefficient processing circuit 23 calculates the average value AVA
It is determined whether or not AVD is greater than a predetermined value t. If an affirmative result is obtained, it is determined whether or not the average value having the smallest value is equal to or less than a predetermined ratio as compared with the other average values. Here, for example, when the average value having the smallest value is equal to or less than 1/2 of the other average values, the coefficient value is set to a value 0 for the block having the smallest average value. As a result, the coefficient processing circuit 23 sets the coefficient to a value of 0 for a block having a degree of coefficient generation equal to or less than a predetermined value as compared to other blocks. At this time, for the block A on the lowest frequency side, even if this condition is satisfied, the process of setting the coefficient to the value 0 is excluded, thereby preventing the image quality deterioration due to the removal of the low frequency component. You.

The coefficient processing circuit 23 calculates the average value AVAＡＶ
If one AVD is less than or equal to the predetermined value t and one of the blocks is other than the block A on the lowest frequency side, the smallest average of the remaining three averages is the other two. It is determined whether the ratio is equal to or less than a predetermined ratio by comparing the two average values. Here, when the average value with the smallest value is, for example, 1/3 or less of the other two average values, in addition to the blocks whose average values AVA to AVD are equal to or less than the predetermined value t, the block with the smallest average value is used. , The coefficient value is set to the value 0. As a result, the coefficient processing circuit 23 determines that there is at least one block on the high frequency side where the degree of occurrence of the coefficient is small, and in the remaining blocks, the degree of occurrence of the coefficient is less than or equal to the predetermined value as compared with the other blocks. , The coefficient is set to the value 0 also for this block. In this case as well, for the block A on the lowest frequency side, even if this condition is satisfied, the process of setting the coefficient to the value 0 is excluded, thereby preventing the image quality deterioration due to the removal of the low frequency component. Is done.

Further, the coefficient processing circuit 23 calculates the average value A of the block D which is the highest frequency side in the horizontal and vertical directions.
When the VD is smaller than the average values AVA to AVC of the other blocks A to C, the coefficient value of the block D is set to a value of zero. Accordingly, when the degree of coefficient generation is smaller for blocks on the higher frequency side in the horizontal and vertical directions than for blocks with lower frequencies in the horizontal or vertical direction, the coefficient processing circuit 23 also performs Is set to the value 0.

By setting the coefficient values, the coefficient processing circuit 23 converts the coefficient data D1 obtained by the discrete cosine transform processing into the coefficient data D1 shown in FIGS.
Processing is performed according to the eight patterns shown by (H). In FIG. 3, blocks indicated by hatching indicate blocks in which coefficient values are not manipulated at all, and blocks not hatched are blocks in which coefficient values are set to 0.

When the coefficient value setting process is completed as described above, or in parallel with the coefficient value setting process, the coefficient processing circuit 23 executes a gain coefficient calculation process. Here, as shown in FIGS. 4A and 4B, the coefficient processing circuit 23 sets the gain coefficient GA having a value of 1 to the block A on the lowest frequency side, and the average value AVA in this block A.
The gain coefficients GA to GD are set such that the values change according to the average values AVA to AVD by solving a predetermined relational expression using the gain coefficients GA and GA.

In this embodiment, the coefficient processing circuit 23 calculates the average value A by using a linear function as the relational expression.
The gain coefficients GA to GD are set so as to be proportional to VA to AVD. As a result, the coefficient processing circuit 23 performs the operation shown in FIG.
In the example shown in FIG. 7, a gain coefficient GA having a value of 1.0 is set for a block A having an average value AVA of 35, and the average value AVB is set to a value of 26 based on the average value AVA and the gain coefficient GA. For block B, a gain coefficient GB having a value of 0.75 is set. Similarly, the average value AV
For blocks C and D with C and AVD values of 17 and 8, gain coefficients GC and GD of values 0.5 and 0.25
Set.

The coefficient processing circuit 23 weights the result of the coefficient value setting processing based on the gain coefficients GA to GD calculated in this manner, and obtains the coefficient data D as the processing result.
2 are output in the order of input. The coefficient processing circuit 23 executes a coefficient value correction process by performing the coefficient value setting process and the coefficient value weighting process, and effectively avoids deterioration of feature points due to the coefficient values, thereby reducing the information amount of the coefficient data.

Further, the coefficient processing circuit 23 calculates the average value dct mavav of the absolute values of the coefficient values in macroblock units for the coefficient data D2 obtained by processing the coefficient values in this manner.
[N] is calculated, and the average value dctmbav [n] is output to the rate control circuit 25 as control data DS.

Inverse discrete cosine conversion circuit 24
Performs inverse discrete cosine transform processing on the coefficient data D2 output from the coefficient processing circuit 23, and outputs the resulting image data to the subtractor 3. Accordingly, in the encoding device 20, the image data input to the subtracter 3 is directly subjected to the discrete cosine transform processing by the discrete cosine transform circuit 5, or the difference data from the predicted frame is subjected to the discrete cosine transform circuit 5 by the discrete cosine transform circuit 5. The conversion processing is performed, and the resulting coefficient data is quantized by the quantization circuit 6.

The rate control circuit 25 includes a coefficient processing circuit 23
Based on the average value dct mav [n] output from the above, the quantization scale value described above with respect to Expression (1) is corrected, and the quantization circuit 6 is corrected by the corrected quantization scale value.
Control the operation of.

That is, the rate control circuit 25 calculates the average value dct m of the absolute values of the coefficient values for each macroblock.
From bav [n], the arithmetic processing of the following equation is executed, whereby the average value dct fr of the absolute values of the coefficient values in frame units
mav is sequentially calculated. Here, N is the number of macro blocks constituting one frame.

[0048]

(Equation 2)

The rate control circuit 25 determines the average value dct for the subsequent frame based on the frame-wise average value dct frmav calculated in this way.
The value of ctmbav [n] is reduced to 1/2 by the following equation.
Normalize in the range of value 2. The rate control circuit 25 sets the normalized value no such that when the average value dct frmav is increased by the normalization process, the value is reduced conversely.
rm_mbav [n] is set.

[0050]

(Equation 3)

The rate control circuit 25 calculates the quantization scale mQ [n] by executing the following equation using the normalized value norm mbav [n] calculated as described above. As a result, the rate control circuit 25 calculates the average value dct frmav based on the coefficient value in the immediately preceding frame.
Increases, the quantization scale mQ [n] is set so that the value decreases in the opposite manner, and this quantization scale mQ
The operation of the quantization circuit 6 is controlled by [n].

[0052]

(Equation 4)

(1-2) Operation of First Embodiment In the above configuration, image data DV1 sequentially input
(FIG. 1), the arrangement of the pictures and the scan order are changed by the preprocessing circuit 2, and the discrete cosine transform circuit 22 performs discrete cosine transform processing. When the image data DV1 is sequentially arranged in frequency order in the horizontal direction and the vertical direction, the image data DV1 is converted into coefficient data D1 in which coefficient values are arranged in a matrix, and the coefficient data D1 is input to the coefficient processing circuit 23.

In the coefficient processing circuit 23, the coefficient data D1 is obtained by dividing the matrix arrangement into 4 × 4
(FIG. 2), and average values AVA to AVD of the absolute values of the coefficient values are calculated for each of the blocks A to D. Here, in the blocks A to D including many features of the DCT block, the average values AVA to AVD calculated in this way show large values.

As a result, the value of the coefficient data D1 is corrected on the basis of the average values AVA to AVD, and the information amount is reduced without deteriorating the characteristic points based on the coefficient values.

At this time, the lowest frequency block A
For, the average value is calculated from the AC coefficient values excluding the 0th-order horizontal and 0th-order coefficients indicating the level of the DC component. For the 0th-order horizontal and 0th-order vertical coefficients indicating the level of the DC component, , The coefficient value is not changed, whereby image quality deterioration due to a change in the DC level is effectively avoided.

That is, in the coefficient data D1, the coefficient value setting processing is first executed in the correction processing of the coefficient data D1, and the coefficient value is set to the value 0 for the block having a small effect on the image quality of the DCT block. .

More specifically, the coefficient data D1 has an average value AV
For blocks where A to AVD are less than or equal to the predetermined value t, all coefficient values are set to the value 0, whereby the coefficients of blocks with a small degree of coefficient generation are set to the value 0. When the average values AVA to AVD are all larger than the predetermined value t, when the average value having the smallest value is equal to or less than a predetermined ratio as compared with the other average values, the block having the smallest average value is selected. , The coefficient value is set to a value of 0, whereby the coefficient value is set to a value of 0 also for a block whose degree of coefficient generation is smaller than a predetermined value as compared with other blocks.

The coefficient data D1 has the average values AVA-A
In the case where one VD is equal to or less than the predetermined value t and one of these is other than the block A on the lowest frequency side, the average value having the smallest value among the remaining three average values is If it is less than or equal to a predetermined ratio compared to the other two averages,
The coefficient value is set to the value 0 also for the block having the smallest average value. Thereby, the coefficient data D1
If there is at least one block on the high frequency side where the degree of occurrence of the coefficient is small, and among the remaining blocks, the degree of occurrence of the coefficient is equal to or less than a predetermined value as compared with the other blocks, the coefficient for this block is also Is set to the value 0.

Further, the coefficient data D1 is the average value AV of the block D which is the highest frequency side in the horizontal and vertical directions.
If D is smaller than the average value AVA-AVC of the other blocks A-C, the coefficient value of this block D is set to the value 0, whereby the block D on the higher frequency side in the horizontal and vertical directions is set. When the degree of coefficient generation is smaller than that of the other blocks, the coefficient is set to the value 0 also for this block D.

In this way, the coefficient data D1 is
Deterioration is effectively avoided for features in blocks, and the amount of information is reduced.

Further, in the coefficient data D1, a gain coefficient GA having a value of 1 is set in the block A, and the value changes in accordance with the average values AVA to AVD based on the average value AVA and the gain coefficient GA in the block A. The gain coefficients GA to GD are set as described above. After the coefficient data D1 has its information amount reduced by the coefficient setting process, it is weighted by the gain coefficients GA to GD calculated in this way, thereby further reducing the information amount.

Thus, in the encoding device 20, the coefficient data D2 obtained by processing the coefficient values in this way is:
The inverse discrete cosine transform circuit 24 performs an inverse discrete cosine transform process to restore the original image data, and this image data is subjected to an intra-frame encoding process and an inter-frame encoding process and output.

At this time, the image data is subjected to a discrete cosine transform process in a discrete cosine transform circuit 5, then quantized by a quantizing circuit 6, and further subjected to a variable length coding process by a variable length coding circuit 8, whereby the data The encoding process is performed with the amount significantly reduced.

At the time of quantization in the quantization circuit 6,
In the coefficient data output from the discrete cosine transform circuit 5, an average value dct mbav [n] of the absolute value of the coefficient data is calculated by the coefficient processing circuit 23 for each macroblock unit, and the average value dctmbav [n] is used as a reference. Then, the quantization scale value is set so that the value is reduced in a macroblock in which the amount of generated coefficient data is large, and quantization processing is performed. As a result, the coefficient data is coded by assigning a large data amount to a portion where the generated code amount is large, thereby effectively avoiding deterioration of the image quality.

That is, a macroblock having a large average value dctmbav [n] tends to generate a large amount of code. On the other hand, if the uniform code amount is assigned by the equation (1), the quantization block value having a large value is set for the macroblock having a large average value by increasing the value of the weight coefficient activity. Deteriorates.

In this embodiment, when the average value dctmbav [n] in a macroblock is increased with reference to the coefficient data in the immediately preceding frame, the value is normalized so as to decrease, and then the normalized value is reduced. By correcting the quantization scale value by the calculated value, a large data amount is allocated to a portion where the generated code amount is large, and the coding process is performed, thereby effectively preventing the deterioration of the image quality.

(1-3) Effects of the First Embodiment According to the above configuration, the average value of the absolute values of the coefficient data is calculated for each block obtained by subdividing the DCT block.
By correcting the coefficient data on the basis of the average value, it is possible to effectively avoid the deterioration of the image quality and efficiently compress the data even for a high-definition video signal.

That is, for a small block whose average value is equal to or less than a predetermined value, the coefficient value is set to a value of 0 and the coefficient value is corrected, thereby effectively avoiding deterioration of image features and reducing the amount of information. As a result, even for a high-definition video signal, it is possible to effectively avoid image quality degradation and efficiently compress data.

The deterioration of image characteristics can also be effectively achieved by setting the coefficient value to a value of 0 for a small block whose average value is equal to or smaller than a certain value as compared with the average value of the other small blocks. Thus, the amount of information can be reduced, and thus, even for a high-definition video signal, the deterioration of the image quality can be effectively avoided and the data can be efficiently compressed.

Further, a weighting coefficient whose value changes in accordance with the average value is generated based on the lowest frequency side minute block, and the coefficient value of each minute block is weighted by the weighting coefficient to correct the coefficient value. By doing so, the amount of information can be further reduced.

The average value of the absolute values of the coefficient values is calculated for each macroblock in units of blocks, and the average value of the absolute values of the coefficient values for the immediately preceding frame is calculated. By correcting the quantization scale value in the quantization means according to the correction result obtained by correcting the average value with the average value in the immediately preceding frame,
A large data amount can be assigned to a portion where the generated code amount is large, and the coding process can be performed. As a result, deterioration of the image quality can be effectively avoided.

Further, in calculating the average value of the absolute values used in these processes, the DC coefficient is excluded from the calculation, and the DC coefficient is processed so that the value does not change. The change can be effectively avoided to prevent the image quality from deteriorating.

(2) Second Embodiment FIG. 5 is a block diagram showing an encoding device according to a second embodiment of the present invention. This encoding device 30 includes a coefficient processing circuit 2 between a discrete cosine transform circuit 5 and a quantization circuit 6 instead of between the preprocessing circuit 2 and the subtractor 3.
Place 3. In the configuration shown in FIG. 5, FIG.
The same components as those of the encoding device 20 described above are denoted by the corresponding reference numerals, and redundant description will be omitted.

That is, in this embodiment, coefficient data obtained by performing discrete cosine transform processing on image data DV1 in intra-frame coding processing, and differential data obtained by performing discrete cosine transform processing on differential data in inter-frame coding processing. The coefficient data is processed and output in the same manner as in the first embodiment. Further, control data DS is generated from the coefficient data and output.

According to the configuration shown in FIG. 5, the same effect as in the first embodiment can be obtained even if coefficient data of data obtained by motion compensation is directly processed and corrected.

(3) Other Embodiments In the above-described embodiment, a case where the array of coefficient data by one DCT block is divided into four blocks and processed is described. The present invention is not limited to this, and may be processed in various blocks as necessary.

In this case, for example, as shown in FIG.
It is also possible to divide the blocks into blocks A to P with a 2 × 2 coefficient array, calculate the average value of the absolute values of the coefficient values, and correct the coefficient values based on the calculation results. Further, the present invention is not limited to the case where blocks are formed by the same number of coefficients in the horizontal and vertical directions, but may be formed by the numbers of coefficients different in the horizontal and vertical directions.

Further, in the above-described embodiment, the case where the gain coefficient is set so as to be proportional to the average value on the basis of the block on the lowest frequency side has been described. However, the present invention is not limited to this. The gain coefficient may be set by calculation using the following function. Further, for example, as shown in FIG. 6B by comparison with FIG. You may give it.

In the above-described embodiment, a case has been described in which the same block is used for the coefficient setting processing and the coefficient weighting processing, but the present invention is not limited to this. Alternatively, in the coefficient weighting processing, the weighting coefficient may be set directly according to the coefficient value without blocking. In this case, by setting the gain coefficient in units of smaller blocks, it is possible to prevent image quality deterioration due to weighting the gain of the coefficient value by the coefficient.

Further, in the above-described embodiment, the case where the coefficient value is set to the value 0 by the constant reference value (t) or the like in the coefficient value setting processing has been described, but the present invention is not limited to this. This criterion may be adaptively switched. In this case, for example, switching based on picture types (I, P, B pictures), switching based on luminance data and chrominance data, and further switching based on a scene change or flash can be considered.

In the above-described embodiment, a case has been described in which the quantization scale value is controlled by calculating the average value of the absolute values from all the coefficient data included in the macroblock. However, the present invention is not limited to this. For example, among the blocks A to D described above with reference to FIG. 2, the average value of the absolute values of the coefficient data included in the macroblock except for the block A on the lowest frequency side may be calculated. By doing so, it is possible to prevent ringing-like distortion (mosquito noise) that is likely to occur in an image containing a high-frequency component.

Further, in the above-described embodiment, the case where the coefficient is corrected and the quantization scale value is also controlled based on the coefficient value has been described. However, the present invention is not limited to this, and the coefficient value may be simply adjusted. Even if the quantization scale value is controlled on the basis of, the image quality can be improved.

[0084]

As described above, according to the present invention, an average value of absolute values of coefficient data is calculated for each block obtained by subdividing a DCT block, and the coefficient data is corrected based on the average value. Even for a high-definition video signal, it is possible to effectively avoid the deterioration of the image quality and efficiently compress the data.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating an encoding device according to a first embodiment of the present invention.

FIG. 2 is a schematic diagram used to explain subdivision of a DCT block in the encoding device of FIG. 1;

FIG. 3 is a schematic diagram illustrating a pattern of a coefficient value setting process.

FIG. 4 is a schematic diagram for explaining a setting process of a gain coefficient;

FIG. 5 is a block diagram illustrating an encoding device according to a second embodiment of the present invention.

FIG. 6 is a schematic diagram used to describe a process performed by an encoding device according to another embodiment.

FIG. 7 is a block diagram showing a conventional encoding device.

FIG. 8 is a chart for explaining coefficient data obtained by discrete cosine processing;

[Explanation of symbols]

1, 20, 30 ... coding device, 5, 22 ... discrete cosine transform circuit, 6 ... quantization circuit, 8 ... variable length coding circuit, 9 ... buffer, 7, 25 ... rate control circuit , 13,24 ... inverse discrete cosine transform circuit, 23 ... coefficient processing circuit

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5C059 KK01 KK27 MA00 MA05 MA23 MC14 ME01 NN21 PP04 SS03 TA43 TA48 TB08 TC03 TC04 TC10 TC12 TC20 TD03 TD12 UA02 5C078 BA57 CA01 DA00 DA01 DA16 DB05 9A001 EE04 EE05 HZ27

Claims (12)

[Claims] An image data which is sequentially input is subjected to a discrete cosine transform process in a predetermined block unit to generate coefficient data in which coefficient values are arranged in a matrix when sequentially arranged in frequency order in a horizontal direction and a vertical direction. Orthogonal transformation means, for each small block obtained by dividing the matrix-like array, and average value calculation means for calculating the average value of the absolute value of the coefficient value, and for each small block based on the average value. An encoding apparatus comprising: a coefficient value correction unit that corrects a coefficient value; and a quantization unit that quantizes the coefficient data corrected by the coefficient value correction unit. 2. The apparatus according to claim 1, wherein said coefficient value correcting means sets the coefficient value to a value of 0 for a small block whose average value is equal to or less than a predetermined value, and corrects the coefficient value. An encoding device according to claim 1. 3. The coefficient value correcting means sets the coefficient value to a value 0 for a small block whose average value is equal to or less than a fixed value compared to the average value of other small blocks, and sets the coefficient value to The encoding device according to claim 1, wherein the encoding is corrected. 4. The coefficient value correcting means generates a weighting coefficient whose value changes in accordance with the average value with reference to the smallest block on the lowest frequency side. The encoding apparatus according to claim 1, wherein the coefficient value is corrected by weighting the coefficient value. 5. A quantization control unit for controlling an operation of the quantization unit, wherein the quantization control unit calculates an average value of absolute values of the coefficient values in units of the block, For a frame, calculate an average value of absolute values of the coefficient values, correct an average value in units of the blocks with an average value in the immediately preceding frame, and quantize the quantization means in accordance with the correction result. The encoding device according to claim 1, wherein the scale value is corrected. 6. An orthogonal transform unit for generating coefficient data by performing discrete cosine transform processing on image data sequentially input in predetermined block units; a quantizing unit for quantizing the coefficient data; And quantization control means for controlling an operation, wherein the quantization control means calculates an average value of absolute values of the coefficient data in units of the block, and for the immediately preceding frame, an absolute value of the coefficient data. Calculating the average value of the blocks, correcting the average value in units of the blocks with the average value in the immediately preceding frame, and correcting the quantization scale value in the quantization means according to the correction result. Encoding device. 7. A discrete cosine transform process for sequentially input image data in a predetermined block unit to generate coefficient data in which coefficient values are arranged in a matrix when sequentially arranged in a frequency direction in a horizontal direction and a vertical direction. And an encoding method for quantizing and encoding the coefficient data, wherein, for each small block obtained by dividing the matrix-like array, an average value calculation processing for calculating an average value of absolute values of the coefficient values; A coefficient value correction process for correcting the coefficient value for each minute block based on the average value. 8. The coefficient value correcting process according to claim 7, wherein the coefficient value is corrected to 0 for a small block whose average value is equal to or less than a predetermined value. Coding method as described. 9. The coefficient value correcting process includes the steps of: setting the coefficient value to 0 for a small block whose average value is equal to or less than a fixed value compared to the average value of other small blocks; The encoding method according to claim 7, wherein the correction is performed. 10. The coefficient value correction processing includes: generating a weighting coefficient whose value changes in accordance with the average value with reference to a minute block on the lowest frequency side; The encoding method according to claim 7, wherein the coefficient value is corrected by weighting the coefficient value. 11. An average value of absolute values of the coefficient values is calculated for each block, an average value of absolute values of the coefficient values is calculated for the immediately preceding frame, and the average value of each block is calculated based on the immediately preceding frame. 8. The encoding method according to claim 7, wherein correction is performed using an average value in a frame, and a quantization scale value in the quantization process is corrected according to the correction result. 12. An encoding method for generating coefficient data by performing discrete cosine transform processing on image data sequentially input in predetermined block units, and performing quantization processing and encoding processing on the coefficient data, wherein the block Calculate the average of the absolute values of the coefficient data in units, calculate the average of the absolute values of the coefficient data for the immediately preceding frame, and calculate the average of the blocks in units of the average in the immediately preceding frame. And a quantization scale value in the quantization process is corrected in accordance with the correction result.