JP2002262294A - System and method for controlling generated code quantity from image signal coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a system for controlling a generated code quantity from an image signal coder that reduces the generated code quantity so as to prevent occurrence of a buffer overflow while avoiding image deterioration even when the coder receives an image with many information quantities such as a fine image pattern and a fast moving image and a target code quantity of a buffer section is small. SOLUTION: The system for controlling the generated code quantity from the image signal coder that divides a frame into blocks, forms a difference signal between a current frame and a preceding frame for each block, converts the difference signal into a frequency component and codes a conversion coefficient, is provided with an in-frame variance calculation section that calculates the variance of the conversion coefficient of an AC component for each coordinate of a block in the frame and cut-off pattern setting sections 3, 23 that leave the conversion coefficients 3, 22 of low frequency parts in the block and the conversion coefficient of the high frequency part with greater variance consecutively in horizontal and vertical directions as they are attending an increase in the generated code quantity and form a cut-off pattern of the coding with respect to the other high frequency conversion coefficients.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal encoding apparatus for efficiently assigning a code amount within a target bit rate range when an image signal is compression-encoded.
In particular, the present invention provides an image signal that enables a stable encoding process without exceeding the target code amount without being able to suppress the generated code amount when the pattern is fine, the motion is large, and the information amount of the input image is large. The present invention relates to a generated code amount control system and method for an encoding device.

[0002]

2. Description of the Related Art FIG. 20 is a block diagram showing a schematic configuration of an image signal encoding apparatus on which the present invention is based. Note that the same components will be denoted by the same reference numerals and symbols throughout the drawings. As shown in the figure, image data is input to the subtraction unit 1 and the motion detection unit 10 of the image signal encoding system.

A motion detecting section 10 stores input image data, detects the distance, direction, etc., of a subject on each screen between a current frame and a previous frame, and forms motion information as a motion vector. . A motion compensation unit 12 is connected to the output side of the motion detection unit 10, and the motion compensation unit 12 attempts to reduce the amount of information in a moving area of the subject. Using the motion information, in the previous frame having the subject before moving,
The subject is shifted and overlapped in accordance with the position after the movement of the current frame to form image data in which the movement of the previous frame is compensated.

In this case, it is necessary to transmit the motion vector to the decoding section. The subtraction unit 1 generates a prediction error image, which is difference data of an image between the input image data and the image data that compensates for the motion of one frame before output from the motion compensation unit 12. As described above, the subtraction unit 1 forms the prediction error image obtained by the motion compensation / inter-frame prediction by setting the difference data to zero even in a portion where there is a motion.
This is to increase the compression ratio by preventing large difference data from being generated.

[0005] A discrete cosine transform unit (DCT unit) 2 is connected to the output side of the subtraction unit 1. The DCT 2 receives the difference data of the image from the subtraction unit 1, that is, the data of the prediction error image, and converts the data into frequency components. And outputs conversion coefficient data. A quantizing unit 5 is connected to the output side of the DCT 2, and the quantizing unit 5 quantizes the transform coefficient data of the DCT 2.

The output side of the quantization unit 5 is connected to a variable length coding unit 6 and an inverse quantization unit 8. The inverse quantization unit 8 performs a process of inversely quantizing the quantized data from the quantization unit 5 to return to the transform coefficient data. However, since there is a component lost by the quantization processing in the quantization unit 5, the transform coefficient returned by the inverse quantization processing does not return to the same transform coefficient data as the output of the DCT2. The lost components are
This is called a quantization error, and this causes deterioration of image quality.

[0007] An inverse DCT 9 is connected to the output side of the inverse quantization unit 8, and the inverse DCT 9 performs inverse discrete cosine transform on the transform coefficient data from the inverse quantization unit 8 to obtain image data. The inverse DCT 9 and the adder 1 are provided on the output side of the motion compensator 12.
4, the adding unit 14 adds the image data for compensating the motion of one frame before and the difference data of the subtracting unit 1,
It is returned to the input image data.

The returned input image data contains the above-mentioned quantization error and is called local decoded data because it is image data of the same image quality as the decoding side. A frame delay unit 13 is connected to the output side of the addition unit 14, and the frame delay unit 13 delays the locally decoded data output from the addition unit 14 by one frame and outputs the data to the motion compensation unit 12.

An encoding control unit 4 is connected to the quantization unit 5 and the inverse quantization unit 8, and the encoding control unit 4 performs coarse quantization and inverse quantization on the quantization unit 5 and the inverse quantization unit 8. It controls generation of a Q (quantization) parameter for performing quantization or fine quantization and inverse quantization. The variable length coding unit 6 receives the quantized data from the quantization unit 5, the motion vector data from the motion detection unit 10, and the Q parameter from the coding control unit 4, and performs variable length coding using Huffman code or the like. Convert to a pre-defined format and output.

[0010] When the motion vector for each small area is encoded, the motion detecting section 10 suppresses the variation of the motion vector and makes the motion vector uniform for each small area. As a result, the appearance frequency is increased, and the amount of coding of the motion vector data is reduced. The variable length encoding unit 6 includes a buffer unit 7
Is connected, and the buffer unit 7 is a speed smoothing memory,
The coded data is efficiently transmitted to the communication line in the order of the previously stored data within the range of the target bit rate.

The encoding control unit 4 connected to the buffer unit 7
Monitors the amount of code stored from the buffer unit 7 and controls generation of the Q parameter. That is, since the transmission of the coded data from the buffer unit 7 is performed with the target code amount, when the generated code amount continues to exceed the target code amount, the accumulated code amount increases. When the state in which the amount is less than the target code amount continues, the accumulated code amount decreases.

For this reason, if an overflow or an underflow occurs in the buffer unit 7 using the difference between the generated code amount and the target code amount as a control signal in the encoding control unit 4, normal encoded data will not be output. When an overflow is likely to occur, the encoding control unit 4 controls the generation of the Q parameter so as to perform coarse quantization and inverse quantization.

Conversely, when an underflow is likely to occur, the encoding control unit 4 controls the generation of the Q parameter so as to perform fine quantization and inverse quantization.

[0014]

However, in the above image signal encoding system, when an image having a large amount of information such as a fine pattern or a large amount of motion is input and the target code amount is small, the Q parameter I can't control it alone,
There is a problem that the buffer unit 7 may overflow.

The reason is that, for example, MPEG (Moti)
on Picture Coding Experts
This is because, in the video coding of Group 2, the upper limit of the Q parameter is determined, and even if the value is quantized, the generated code amount may not be smaller than the target code amount. If the generated code amount continues to be larger than the target code amount, an overflow occurs because the data amount input to the buffer unit 7 is larger than the data output from the buffer unit 7. then,
In the buffer unit 7, since newly input data is overwritten by past data that has not been output yet, the encoded data output from the device does not satisfy the prescribed format and is corrupted. It becomes data.

Therefore, in view of the above problems, the present invention avoids image degradation even when an image having a large amount of information such as a fine pattern or a large amount of motion is input and the target code amount of the buffer section is small. It is another object of the present invention to provide an image signal encoding system and method capable of reducing the amount of generated codes and preventing occurrence of overflow in a buffer unit.

[0017]

In order to solve the above-mentioned problems, the present invention divides a frame into a plurality of blocks, forms a difference signal between a current frame and a previous frame for each block, and generates the difference signal. In a generated code amount control system of an image signal encoding device that encodes a transform coefficient by converting to a frequency component, an intra-frame variance calculation unit that calculates a variance of an AC component transform coefficient for each coordinate of the block in the frame. ,
With the increase in the generated code amount, the transform coefficients of the low-frequency portion in the block and the transform coefficients of the high-frequency portion having a large variance continuously in the horizontal and vertical directions are left as they are, and other high-frequency transforms are performed. And a truncation pattern setting unit that forms a truncation pattern for encoding of coefficients.

By this means, even when an image having a large amount of information such as a fine pattern or a large amount of motion is input and the target code amount of the buffer section is small, the generated code amount can be reduced while avoiding image deterioration. However, it is possible to prevent the buffer unit from overflowing. Preferably, the termination pattern setting unit has the termination pattern as a matrix in advance.

By this means, the process of cutting the high-frequency coding amount can be performed quickly. Preferably, the truncation pattern setting unit gradually increases the truncation pattern from a high band to a low band as the generated code amount increases. By this means, the process of cutting the high-frequency coding amount can be performed quickly while avoiding image deterioration.

Preferably, the conversion of the difference signal into a frequency component is a DCT or Hadamard transform. By this means, the present invention can be applied to a general image signal encoding device. Preferably, the difference signal is formed from the current frame and the previous frame that has been motion-compensated, and the motion vector of the subject between the current frame and the previous frame is encoded together with the difference signal. The coding of the motion vector is terminated.

By this means, it is possible to further prevent the overflow of the buffer unit by cutting off the coding of the motion vector together with the high frequency cut of the transform coefficient. Preferably, as the generated code amount increases, the transform coefficients are coarsely quantized or dequantized, coding of high-frequency transform coefficients is stopped, and coding of motion vectors for performing motion compensation is stopped. . By this means, the amount of generated codes can be reduced more efficiently.

Further, according to the present invention, a frame is divided into a plurality of blocks, a difference signal between a current frame and a previous frame is formed for each block, and the difference signal is converted into a frequency component to encode a transform coefficient. In the generated code amount control method of the image signal encoding device, a step of calculating a variance of a transform coefficient of an AC component for each coordinate of the block in a frame; and, with an increase in the generated code amount, a low frequency band in the block. Transform coefficients of the portion, and leaving the transform coefficients of the high-frequency portion having a large variance continuously in the horizontal direction and the vertical direction as they are, and forming a truncation pattern of the encoding for the other high-frequency transform coefficients. And a method for controlling a generated code amount of the image signal encoding apparatus.

By this means, similar to the above-described invention, even when an image having a large amount of information such as a fine pattern or a large amount of motion is input and the target code amount of the buffer section is small, image deterioration can be avoided. , The amount of generated codes can be reduced, and the occurrence of overflow in the buffer unit can be prevented.

[0024]

Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of a generated code amount control system of an image signal encoding device according to the present invention. As shown in this figure, the components different from those in FIG. 4 are a DCT adjustment unit 3 and a vector adjustment unit 11.

The DCT adjustment unit 3 is connected between the subtraction unit 1 and the quantization unit 5 and receives a coefficient cut control signal from the coding control unit 4. Further, the DCT adjustment unit 3 cuts the high frequency component of the DCT coefficient from the DCT 2 according to the coefficient cut control signal from the encoding control unit 4, removes the high frequency component, and reduces the generation of the code amount of the DCT coefficient. enable. At this time, if the generated code amount in the buffer unit 7 is small with respect to the target code amount and there is room, a process of reducing or not cutting the amount of cutting the high frequency component of the DCT coefficient is performed. If the generated code amount in the buffer unit 7 is relatively large with respect to the target code amount, a process of increasing the amount of cutting the high frequency component of the DCT coefficient is performed.

Next, the vector adjustment section 11 is
0 and a motion vector control signal from the coding control unit 4. Further, the vector adjustment unit 11 adjusts the motion vector output from the motion detection unit 10 according to the motion vector control signal from the encoding control unit 4. At this time, when the generated code amount in the buffer unit 7 is small with respect to the target code amount and there is a margin, a process of reducing or not adjusting the amount of adjustment is performed. When the generated code amount in the buffer unit 7 is relatively large, a process of increasing the amount to be adjusted is performed.

By adjusting this motion vector,
The subtraction unit 1 forms a prediction error image obtained by inter-frame prediction without motion compensation. For this reason, if only the motion vector is adjusted alone, large difference data in a portion having a motion cannot be reduced. But,
By cooperating with the DCT adjusting unit 3, the DCT adjusting unit 3 cuts high frequency components of the prediction error image and cuts large difference data of a portion where there is motion, so that the effect of motion compensation is lost.

In such a case, the vector adjustment unit 11
Since it is no longer necessary to encode and transmit the motion vector via the variable length encoding unit 6, the motion vector is adjusted to reduce the amount of code of the motion vector. FIG. 2 is a block diagram showing a schematic configuration of the DCT adjustment unit 3 in FIG. As shown in the figure, the DCT adjustment unit 3 is provided with a cutoff amount setting unit 21. The cutoff amount setting unit 21 receives a coefficient cut control signal from the encoding control unit 4 and determines the degree of high frequency cut. Set according to the coefficient cut control signal.

That is, in the censoring amount setting unit 21, if the generated code amount in the buffer unit 7 continues to be larger than the target code amount and the buffer unit 7 is likely to overflow, the censoring degree is increased and the DCT is increased. An operation of suppressing the generated code amount of the coefficient is performed. Conversely, in the censoring amount setting unit 21, the state in which the amount of generated code in the buffer unit 7 is smaller than the target amount of code continues, and if the buffer unit 7 is likely to underflow, the degree of censoring is reduced. Alternatively, the generated code amount of the DCT coefficient is not suppressed by not performing the truncation.

As described above, the cutoff amount setting unit 21 auxiliary controls the generated code amount of the DCT coefficient. Further, the DCT adjusting unit 3 is provided with an intra-frame variance calculating unit 22.
Input the coefficient data 1 of the DCT coefficient. That is,
In DCT2, a frame is divided into, for example, 8 × 8 pixels, and DCT coefficients are calculated for each block by DCT2. Each block has 64 DCT coefficients. The unit 22 calculates a variance value for each DCT coefficient of each coordinate in each block in the frame.

A discontinuation pattern setting unit 23 is connected to the intra-frame variance calculation unit 22, which receives the variance value data from the intra-frame variance calculation unit 22 according to the input variance value data. , A pattern of what kind of high frequency cut is to be performed. In the intra-frame variance calculation unit 22, for example, in a frame having a large number of horizontal frequency components, if the horizontal frequency components are blindly cut, the image quality deteriorates accordingly. Therefore, in a frame having many frequency components in the horizontal direction, a process of preferentially cutting the frequency components in the vertical direction is performed.

A control signal generator 24 is connected to the censoring amount setting unit 21 and the censoring pattern setting unit 23. The signal is used to form a censoring control signal. The DCT adjusting unit 3 is provided with a 0-substitution unit 25. The 0-substitution unit 25 is configured to output the coefficient data 1 of the DCT coefficient input from the DCT2 in accordance with the truncation control signal generated by the control signal generation unit 24.
, The DCT coefficient of the corresponding coordinate is replaced with “0” and output as coefficient data 2.

In this case, the DCT coefficient replaced with "0" is not transmitted to the decoding side, and the generated code amount is reduced accordingly. From the general nature of the image, the higher the high-frequency component, the more likely the variance is to be small.By uniformly cutting the high-frequency component, it is possible to reduce the amount of generated code. Due to the movement, the high-frequency component increases, and if this is uniformly cut, the image quality deteriorates. For this reason, it is necessary to perform high-frequency cut so that the deterioration of the image quality is not increased even for a fine pattern or movement.

Hereinafter, the operation of the DCT adjusting section 3 will be described in detail. FIG. 3 is a diagram illustrating coefficient data 1 of DCT coefficients input from the DCT 2 to the DCT adjustment unit 3. As shown in the figure, in the frame, as an example, 8 × 8 64
The number of pixels is one block, the frame is divided into a plurality of blocks, and coefficient data 1 is formed for each pixel of each block.

The coefficient data 1 of each block is a single direct current (DC) component having a positive value, and has both positive and negative values 63
It is composed of two alternating current (AC) components. It is generally said that the occurrence rate of the AC component of the coefficient data 1 has a Gaussian distribution centered on “0”. FIG. 4 is a diagram for explaining the operation of the intra-frame variance calculation unit 22 of the DCT adjustment unit 3, and FIG. 5 is a diagram showing the variance calculated by the intra-frame variance calculation unit 22. As shown in FIG. 4, the intra-frame variance calculation unit 22 forms a frequency distribution of the coefficient data 1 for each block coordinate, and calculates the variance σij.

As shown in FIG. 5, the intra-frame variance calculator 22 of the DCT adjuster 3 calculates the variance of 63 AC components for all blocks in the frame. Regarding the calculated variance value, the variance is small for a certain AC component and large for a certain AC component. A small variance means that the coefficient components are concentrated near “0”. Conversely, a large variance means that a coefficient value having a size that cannot be ignored exists at a certain frequency.

FIG. 6 is a diagram for explaining an operation for forming a pattern of a large or small variance determined by the termination pattern setting unit 23 of the DCT adjusting unit 3. As shown in the figure, the censoring pattern setting unit 23 determines whether or not σij is greater than a predetermined value σ0 with respect to the coordinates of the high-frequency portion BH of the block. If the variance is large, it is set to “1”. If it is smaller, a truncation pattern of "0" is formed.

The “0” of the censoring pattern is replaced by the 0 substitution unit 25.
And the high frequency cut is performed for the corresponding coefficient data 1. With respect to the coordinates in a certain range (4 × 4 pixels) of the low-frequency portion BL of the block, no truncation pattern is formed. This is because, if the low-frequency coefficient data 1 is to be cut, image deterioration becomes remarkable. Note that the range of the low-frequency portion BL can be further reduced as long as image degradation does not occur.

In the high-frequency portion BH of the block, a high-frequency cut can be efficiently performed by detecting a "0" discontinuation pattern that specifies a small variance.
Next, when it is necessary to further suppress an increase in the generated code amount, it is required to cut even the coefficient data 1 having a large variance in a high frequency band. In this case, a large variance is biased toward a horizontal component or a vertical component due to a pattern or motion, and if this coefficient is ignored and blindly cutting coefficient data 1 having a large variance in a high frequency range, the image quality is degraded. Will do.

For this reason, the cutoff pattern setting unit 23 has a matrix of high frequency cut patterns,
By multiplying the coefficient data 1 by the “0” value of the corresponding matrix, high-frequency cut can be quickly performed as follows. FIG. 7 is a diagram showing an example in which coefficient data having a large variance is cut in the pattern of FIG.

When large variances are arranged in the horizontal and vertical directions, that is, in the high-frequency portion BH of the block in FIG.
For example, when four “1” s are arranged in a row, as shown in the figure, the portion is left as “1”, and all other “1” portions are “0” censoring patterns. To
That is, if there is a horizontal or vertical deviation in the large variance existing in the block except for the low-frequency part BL, the coefficient data 1 having a large variance in the non-biased part except for this deviation is prioritized. Is cut.

Thus, for fine patterns and movements,
High-frequency coefficient data 1 efficiently without deteriorating image quality
Can be cut, and the generated code amount can be suppressed. FIG. 8 is a diagram showing another example in a case where coefficient data having a large variance is cut in the block of FIG.

As shown in this figure, as compared with FIG. 7, the high-frequency portion BH of the block is changed from the high side to the high side BH1, BH
If the pattern is divided into two and four “1” s are continuously arranged in the pattern of the high-frequency part BH, that part is left as “1”. For one of the two high-frequency parts, “1” is uniformly changed to a “0” censoring pattern in the high-side BH1 pattern, and “1”,
"0" is left as it is.

In this manner, it is possible to efficiently cut the high-frequency coefficient data 1 without incurring further deterioration of the image quality for a fine picture or movement, thereby suppressing the generated code amount. It becomes possible to do. FIG. 9 is a diagram showing another example in which coefficient data having a large variance is cut in the pattern of FIG.

As shown in FIGS. 7A to 7E, as compared with FIG. 7, the high-frequency portion BH is shifted from the high side to the high side BH1,
When the pattern is divided into four into BH2, BH3, and BH4, and four "1" s are continuously arranged in the pattern of the high-frequency part BH, that part is left as "1". As shown in FIG.
Regarding the divided high-frequency portion, in the high-side BH1 block, “1” is uniformly changed to a “0” censoring pattern,
In the blocks on the high side BH2 to BH4, "1" and "0" are left as they are.

As shown in FIG. 4B, in the high-frequency portion divided into four, the blocks on the high side BH1 and BH2 are:
"1" is uniformly set to the discontinuation pattern of "0", and "1" and "0" are left as they are in the high-side BH3 and BH4 blocks. As shown in FIG. 9C, regarding the high-frequency portion divided into four, the blocks on the high side BH1 to BH3 are:
“1” is uniformly set to a censoring pattern of “0”, and “1” and “0” are left as they are in the high-side BH4 block.

As shown in FIG. 5E, the high-side blocks divided into four parts are divided into the high-side blocks BH1 to BH4.
“1” is uniformly set to a discontinuation pattern of “0”. In this way, a plurality of high-frequency cut pattern matrices are provided in advance in the censoring pattern setting unit 23, and high-frequency cut can be performed by multiplying the coefficient data 1 by the “0” value of the corresponding matrix. Will be possible.

Further, by performing the function processing, it becomes possible to gradually change the pattern of "1" into the discontinuation pattern of "0" in the high-frequency portion BH of the block. FIG. 10 is a diagram illustrating a coefficient cut control signal output from the encoding control unit 4 in FIG. As shown in the figure, in the coding control unit 4, the difference between the generated code amount input from the buffer unit 7 and the target code amount is formed as a coefficient cut control signal, and is output to the DCT adjustment unit 3.

FIG. 11 is a diagram for explaining the operation of the control signal generator 24 in the DCT adjuster 3. As shown in this figure, the control signal generation unit 24 inputs the cutout amount signal from the cutoff amount setting unit 21 when the threshold value of the coefficient cut control signal is exceeded. In the control signal generation section 24, before the cut amount signal is input from the cut amount setting section 21 while exceeding the threshold value of the coefficient cut control signal, the high-frequency portion BH formed in FIG.
The 0 replacement unit 25 replaces the coefficient data 1 with “0” with the truncation pattern of “0” having a small variance in the block of “1”.

Further, in the control signal generating section 24, after the cut-off amount signal is inputted from the cut-off amount setting section 21 while exceeding the threshold value of the coefficient cut control signal, the block of the high-frequency portion BH formed in FIG. In the case where four “1” s are arranged in a row, as shown in this drawing, the portion is left as “1”, and all other “1” portions are replaced with “0” by the censoring pattern of “0”. The unit 25 replaces the coefficient data 1 with “0”.

In this case, as shown in FIG. 8, the high band portion BH of the block is shifted from the high side to the high sides BH1 and BH2.
When four “1” s are arranged in a row in the pattern of the high-frequency portion BH, the portion is left as “1” and the pattern of the high-side BH1 is divided into one of the two high-frequency portions. In this case, “1” is uniformly changed to a censoring pattern of “0”, and “1”, “0” is used in the other high-side BH2 block.
May be left as it is, and the 0 replacement unit 25 may replace the coefficient data 1 with “0” using such a truncation pattern of “0”.

FIG. 12 is a diagram for explaining a modification of the operation of the control signal generator 24 in the DCT adjuster 3. As shown in the figure, when the control signal generation unit 24 exceeds the threshold values a, b, c, and d of the coefficient cut control signal, four “1” s are continuously arranged in the pattern of the high-frequency portion BH. In
That part is left at "1" and the other parts are
(A), (b), (c), and (d) are the censoring patterns a, b, c, and d, which are “0” censoring patterns.
The 0 replacement unit 25 replaces the coefficient data 1 with “0”.

FIG. 13 shows the vector adjustment unit 11 in FIG.
FIG. 2 is a block diagram showing a schematic configuration of the embodiment. As shown in the figure, the vector adjustment unit 11 is provided with a 0 replacement unit 31 and a cutoff amount setting unit 32. The cutoff amount setting unit 32 receives the motion vector control signal from the encoding control unit 4 and forms a cutoff amount signal.

The 0-substitution unit 31 receives the cut-off amount signal from the cut-off amount setting unit 32 from the encoding control unit 4, receives the motion vector data 1 from the motion detection unit 10, and converts the motion vector data 1 based on the cut-off amount signal. Replace with "0",
The motion vector data 2 excluding the motion vector data 1 replaced with “0” is output to the variable-length encoding unit 6, encoded and transmitted.

Motion vector data 1 replaced with "0"
Is not encoded, the generated encoding amount is reduced. FIG. 14 is a diagram illustrating the motion vector data 1 input from the motion detection unit 10 to the vector adjustment unit 11. As shown in the figure, in the motion detection unit 10, as an example, the inside of a frame is divided into blocks of 16 × 16 pixels.

FIG. 15 is a diagram for explaining an example in which the motion detector 10 detects a motion vector. As shown in the figure, the block of the previous frame is extracted within a certain range around the position of the block of the current frame, and the prediction error is calculated. A vector Vij is determined.

FIG. 16 is a diagram for explaining the motion vector of the subject between the current frame and the previous frame. As shown in the figure, Vij and V of blocks constituting a plurality of moving subjects
kl, Vmn, and Vop are motion vector data 1,
The motion is output from the motion detection unit 10 to the vector adjustment unit 11.
FIG. 17 is a diagram illustrating a motion vector control signal output from the encoding control unit 4 in FIG.

As shown in the figure, the encoding control unit 4
The difference between the generated code amount input from the buffer unit 7 and the target code amount is formed as a motion vector control signal, and is output to the vector adjustment unit 11. FIG. 18 is a diagram illustrating the operation of the discontinuation setting unit 32 in the vector adjustment unit 11. As shown in the figure, when the threshold value of the motion vector control signal is exceeded, the termination setting unit 32 forms a termination signal from the termination setting unit 21, terminates encoding of the motion vector, and adjusts the motion vector.

FIG. 19 is a diagram for explaining the relationship among the control signal of the Q parameter, the coefficient cut control signal, and the control signal of the motion vector. The control signal of the Q parameter, the coefficient cut control signal, and the control signal of the motion vector are formed by the difference signal between the generated code amount and the target code amount, and the threshold of the control signal of the Q parameter <the threshold of the coefficient cut control signal <the control of the motion vector. There is a signal threshold relationship.

That is, when an overflow is likely to occur, first, the generation of the Q parameter is controlled so as to perform coarse quantization and inverse quantization. Controls the coefficient cut, and furthermore, if the control of the coefficient cut does not sufficiently suppress the occurrence of overflow, the motion vector is adjusted.

In this way, the amount of generated codes can be efficiently reduced. In the above description, the present invention is applied to an encoding device that combines the inter-frame difference and the DCT. However, the present invention is not limited to this. The inter-field difference and the device using other orthogonal transform such as the Hadamard transform are also applicable. Applicable. In addition, intra-frame coding, not only inter-frame coding, but also intra-frame coding, a device using inter-field coding, and inter-frame coding, inter-field coding, forward prediction using the past screen, Any combination of backward prediction using a future screen or bidirectional prediction using both past and future screens is applicable.

[0062]

As described above, according to the present invention,
Calculates the variance of the AC component transform coefficient for each block coordinate in the frame, and with the increase in the amount of generated code, continuously increases the transform coefficient of the low-frequency part in the block and the horizontal and vertical variances. Since the transform coefficients of the high-frequency portion having the high-frequency portion are left as they are, and the truncation pattern of the encoding for the transform coefficients of the other high frequencies is formed, an image having a large amount of information such as a fine pattern and a large amount of motion is input. In addition, even when the target code amount of the buffer unit is small, it is possible to reduce the generated code amount and prevent the buffer unit from overflowing while avoiding image degradation.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a schematic configuration of a generated code amount control system of an image signal encoding device according to the present invention.

FIG. 2 is a block diagram showing a schematic configuration of a DCT adjustment unit 3 in FIG.

FIG. 3 shows a DCT input from a DCT 2 to a DCT adjustment unit 3
FIG. 9 is a diagram illustrating coefficient data 1 of a coefficient.

FIG. 4 is a diagram for explaining the operation of the intra-frame variance calculation unit 22 of the DCT adjustment unit 3;

FIG. 5 is a diagram showing the variance calculated by the intra-frame variance calculation unit 22.

FIG. 6 shows a discontinuation pattern setting unit 23 of the DCT adjustment unit 3.
FIG. 7 is a diagram for explaining an operation of forming a large / small pattern of variance determined according to FIG.

FIG. 7 is a diagram illustrating an example of a case where coefficient data having a large variance is cut in the pattern of FIG. 6;

8 is a diagram showing another example in a case where coefficient data having a large variance is cut in the block of FIG. 7;

FIG. 9 is a diagram showing another example in the case of cutting coefficient data having a large variance in the pattern of FIG. 7;

FIG. 10 is a diagram illustrating a coefficient cut control signal output from an encoding control unit 4 in FIG. 1;

FIG. 11 shows a control signal generator 24 in the DCT adjuster 3.
It is a figure explaining operation of.

FIG. 12 shows a control signal generator 24 in the DCT adjuster 3.
FIG. 9 is a diagram illustrating a modification of the operation of FIG.

13 is a block diagram illustrating a schematic configuration of a vector adjustment unit 11 in FIG.

FIG. 14 is a diagram illustrating motion vector data 1 input from the motion detection unit 10 to the vector adjustment unit 11.

FIG. 15 is a diagram illustrating an example in which a motion vector is detected in the motion detection unit 10.

FIG. 16 is a diagram illustrating a motion vector of a subject between a current frame and a previous frame.

17 is a diagram illustrating a motion vector control signal output from the encoding control unit 4 in FIG.

FIG. 18 is a diagram for explaining the operation of the cutoff amount setting unit 32 in the vector adjustment unit 11;

FIG. 19 is a diagram illustrating the relationship among a control signal of a Q parameter, a coefficient cut control signal, and a control signal of a motion vector.

FIG. 20 is a block diagram illustrating a schematic configuration of an image signal encoding device that is a premise of the present invention.

[Explanation of symbols]

 REFERENCE SIGNS LIST 1 subtraction unit 2 DCT 3 DCT adjustment unit 4 encoding control unit 5 quantization unit 6 variable length encoding unit 7 buffer unit 8 inverse quantization unit 9 inverse DCT 10 motion detection unit 11 ... Vector adjustment unit 12 ... Motion compensation unit 13 ... Frame delay unit 14 ... Addition unit 21 ... Stop amount setting unit 22 ... Intra-frame variance calculation unit 23 ... Stop pattern setting unit 24 ... Control signal generation unit 25, 31 ... 0 replacement unit 32 ... Discontinuation setting section

 ────────────────────────────────────────────────── ─── Continued on the front page F term (reference) 5C059 KK35 MA00 MA05 MA23 MA24 MC11 MC38 ME02 NN21 PP04 RC16 RC38 RE11 TA46 TA49 TA61 TB13 TC04 TC12 TD04 UA02 UA32 5J064 AA02 BA01 BA09 BA16 BC01 BC16 BD01

Claims (7)

[Claims] 1. An image signal encoding method for dividing a frame into a plurality of blocks, forming a difference signal between a current frame and a previous frame for each block, converting the difference signal into a frequency component, and encoding a transform coefficient. In the generated code amount control system of the apparatus, an intra-frame variance calculation unit that calculates a variance of a transform coefficient of an AC component for each coordinate of the block in the frame, and a low-frequency band in the block as the generated code amount increases. A transform coefficient of a portion, a horizontal direction, a transform pattern of a high-frequency portion having a large variance continuously in the vertical direction, leaving the transform coefficient as it is, and a truncation pattern setting unit for forming a truncation pattern of encoding for the transform coefficients of other high frequencies. A generated code amount control system for an image signal encoding device, comprising: 2. The generated code amount control system for an image signal encoding device according to claim 1, wherein said censoring pattern setting section has said censoring pattern as a matrix in advance. 3. The image signal encoding apparatus according to claim 1, wherein said censoring pattern setting unit gradually increases the censoring pattern from a high frequency to a low frequency as the generated code amount increases. The generated code amount control system of the device. 4. The method according to claim 1, wherein the conversion of the difference signal into a frequency component is DC.
The generated code amount control system for an image signal encoding device according to claim 1, wherein the system is a T-transform or a Hadamard transform. 5. The difference signal is formed from the current frame and the previous frame subjected to motion compensation, and a motion vector of a subject between the current frame and the previous frame is encoded together with the difference signal. 2. The system according to claim 1, wherein the coding of the motion vector is terminated. 6. A transform coefficient is roughly quantized or inversely quantized with an increase in the generated code amount, and coding of a high-frequency transform coefficient is stopped, and coding of a motion vector for performing motion compensation is stopped. 2. The generated code amount control system for an image signal encoding device according to claim 1, wherein: 7. An image signal coding method for dividing a frame into a plurality of blocks, forming a difference signal between a current frame and a previous frame for each block, converting the difference signal into a frequency component, and coding a transform coefficient. Calculating a variance of a transform coefficient of an AC component for each coordinate of the block in a frame; and transforming a transform coefficient of a low-frequency portion in the block with an increase in the generated code amount. And a step of leaving a transform coefficient of a high-frequency portion having a large variance continuously in the horizontal direction and the vertical direction as it is, and forming a truncation pattern of encoding for other high-frequency transform coefficients. A generated code amount control method for an image signal encoding device.