JPH04259182A - Movement compensation prediction inter-frame coder - Google Patents

Abstract

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

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a motion-compensated predictive interframe coding apparatus for television signals.

0002

[Background Art] In recent years, with the development of video encoding technology, video telephones, video conference systems, CD-ROMs,
A motion compensated predictive interframe coding device has been developed as a highly efficient coding device for color moving images used in digital VTRs and the like. For example, Takahiko Fukinuki, "Multidimensional signal processing of TV images" (November 15, 1988, Nikkan Kogyo Shimbun, Chapter 7 High Efficiency Coding, pp213-pp2)
The motion compensated predictive interframe coding device described in 91) is known.

[0003] In a motion compensated predictive interframe coding device,
In order to achieve video encoding at a constant frame rate, when the amount of generated code is large, the prediction error or the quantization step size of the pixel value of the input television signal is increased to limit the amount of generated code. . As a conventional method for determining the quantization step size, the C.C.I.T.T. document "Description of Reference Model 8" (C.C.I.T.T. SG
XV Document 25 ” title:Descript
ion of Ref. Model 8 (RM8),
source: Working Party XV
/4 Specialist Group On Co
ding for Visual Telephony
, version: June. 9.1989'') is known.

A conventional motion compensated predictive interframe coding apparatus will be described below with reference to FIG. In FIG. 2, 51 is an input terminal to which an input television signal is input;
Reference numeral 3 denotes a motion vector calculation unit that compares the image signal of the encoded block of the current frame with the reproduced image signal of the previous frame to calculate the motion vector of the encoded block, and 54 stores the reproduced image signals of the current frame and the previous frame. An image memory unit, 58 is a motion compensation prediction unit that performs motion compensation prediction on the reproduced image signal of the previous frame, and an interframe/intraframe determination unit that determines whether a block to be encoded is to be intraframe encoded or interframe encoded. , 62 is an in-loop filter unit that performs two-dimensional low-pass filter processing on the motion-compensated prediction signal, 64 is a prediction error calculation unit that calculates a prediction error by calculating the difference between the original image signal and the prediction signal of the encoded block, and 66 is a code 68 is an orthogonal transformation unit that orthogonally transforms the signal to be orthogonally transformed; 70 is a quantizer for orthogonal transformation coefficients; 73 is a quantization step size calculation unit that calculates a quantization step size, 74 is a code memory unit that temporarily stores transmission frames, and 76 is an inverse orthogonal transform unit that performs inverse orthogonal transform on quantized orthogonal transform coefficients. ,7
8 is a reproduced image calculation unit that calculates a reproduced image of the current frame;
82 is a prediction error encoding unit that encodes the prediction error through a channel;
84 is a motion vector encoding unit that encodes a motion vector through a channel; 86 is a multiplexer unit that configures a transmission frame from a prediction code and a motion vector code; and 89 is an output terminal that outputs a transmission signal.

The operation of the above configuration will be explained below. First, a television signal that is converted into a digital signal by an analog-to-digital conversion circuit (not shown) and divided into blocks of M pixels in the horizontal direction and N lines in the vertical direction is input as an input television signal 52 from an input terminal 51. .

The motion vector calculation section 53 compares the input television signal 52 with the reproduced television signal 55 of the previous frame stored in the image memory section 54, calculates the motion of the encoded block as a motion vector, and calculates the motion of the encoded block as a motion vector. The vector signal 56 is output. At the same time, the motion vector calculation unit 53 uses the evaluation value at the time of motion vector calculation to determine whether motion compensation prediction is valid or invalid for the encoded block, and outputs the result as a motion compensation prediction control signal to the motion vector signal 56. . Therefore, the motion vector signal 56 has a motion vector and a motion compensated prediction signal superimposed thereon.

The motion compensation prediction unit 58 (1) performs motion compensation prediction on the reproduced television signal 55 of the previous frame using a motion vector when the motion compensation prediction control signal indicates the validity of motion compensation prediction; 2) If the motion compensation prediction control signal instructs invalidation of motion compensation prediction, the reproduced television signal 55 of the previous frame is output as is as the motion compensation prediction signal 59.

The inter-frame/intra-frame determining unit 60 compares the input television signal 52 and the motion-compensated prediction signal 59 on a block-by-block basis, and determines the validity/invalidity of the motion-compensated prediction. If the effectiveness of (
2) If the effectiveness of motion compensation prediction is high, it is determined that interframe coding is effective for the corresponding block, and the result is output as the coding method selection signal 61. By switching the encoding method between the intra-frame encoding method and the inter-frame encoding method on a block-by-block basis, the following improvements can be made compared to the case where encoding is performed using only the inter-frame encoding method. (1) When a scene change occurs, intra-frame encoding is selected, so the image quality after the scene change can be improved. (
2) When a large movement of the moving object occurs, a background area hidden behind the moving object appears, and in this case, intra-frame encoding is selected, so that the image quality can be improved.

[0009] The in-loop filter section 62 performs two-dimensional low-pass filter processing on the coded block that has been motion compensated and predicted using the motion vector, and calculates a prediction signal 63. The prediction error calculation unit 64 performs a difference calculation between the input television signal 52 and the prediction signal 63 of the encoded block,
The result is output as a prediction error signal 65.

The switch unit 66 (1) selects the input television signal 52 as the signal 67 to be orthogonally transformed when the encoding method selection signal 61 selects intraframe encoding, and (2) selects the input television signal 52 as the signal 67 to be orthogonally transformed. When the method selection signal 61 selects interframe coding, the prediction error signal 65 is selected as the signal 67 to be orthogonally transformed.

The orthogonal transform section 68 converts the signal 67 to be orthogonally transformed.
The high correlation between neighboring pixels of the signal 67 to be orthogonally transformed is removed, and the orthogonal transform coefficient 6
Calculate 9. As the orthogonal transform method, in many cases, discrete cosine transform is used, which has high transform efficiency and can be implemented in hardware.

The quantization unit 70 has a quantization step size of 7
1 to quantize the orthogonal transform coefficient 69 and calculate the orthogonal transform quantized coefficient 72.

The quantization step size calculation unit 73 calculates the quantization step size 71 from the residual code amount 75 in the code memory unit 74 using the method shown below.

In this conventional example, the input television signal has a size of 352 pixels in the horizontal direction and 288 lines in the vertical direction. Macro Blo
ck)". ). The quantization step size Qb is calculated from the equation shown in equation (1) at the start of quantization with a cycle of n macroblocks.

Qb=2×INT[Bcont÷200q]+2
(1)
However, in equation (1), it is defined as follows.

(a) INT[*] is a function that truncates the fractions below the decimal point. (b) Bcont indicates the remaining code amount in the code memory section 74.

(c) q is an encoding speed parameter and has a relationship with the encoding speed V as shown in equation (2).

[0018] V=q×64kbit/sec
(2)
As is clear from equation (1), the residual code amount Bcont
As the number increases, the quantization step size Qb increases, the amount of generated code is limited, and video signal encoding at a constant frame rate can be realized. For example, quantization step size Q
When calculating b, when the residual code amount Bcont=700 bits, the quantization step size Qb=8, and when the residual code amount Bcont=6100 bits, the quantization step size Qb=62.

In this conventional example, the quantization step size Qb
The calculation period n is set to n=12.

The inverse orthogonal transform section 76 performs inverse orthogonal transform on the orthogonal transform quantization coefficients 72 to calculate an orthogonal transform signal 77 containing a quantization error.

The switch section 66 (1) selects the numerical value "0" signal 80 as the reproduction image calculation signal 79 when the encoding method selection signal 61 selects intraframe encoding, and (2) selects the numerical value "0" signal 80 as the reproduction image calculation signal 79. When the coding method selection signal 61 selects interframe coding, the prediction signal 63 is selected as the reproduced image calculation signal 79.

The reproduced image calculation unit 78 adds the orthogonal transform signal 77 containing the quantization error and the reproduced image calculation signal 79 to calculate the reproduced image signal 81 of the encoded block.

The image memory 54 stores the reproduced image signal 81 of the current frame and outputs the reproduced image signal 55 of the previous frame.

The prediction error encoding unit 82 encodes the orthogonal transform quantization coefficient 72, the quantization step size 71, and the encoding method selection signal 61, and calculates a prediction error code 83. The quantization step size 71 is encoded only when the value of the quantization step size 71 changes, that is, once every n macroblocks.

The motion vector encoder 84 encodes the motion vector 56
is encoded, and a motion vector code 85 is calculated.

The multiplexer unit 86 uses the prediction error code 83
A transmission frame 87 in a predetermined format is calculated from the motion vector code 85.

The code memory section 74 stores the transmission frame 87 as
Once accumulated, the signal is output from an output terminal 89 as a transmission code 88 in synchronization with a clock signal input from an external source (not shown). At the same time, the code memory unit 74 calculates the amount of codes remaining in the memory as the remaining code amount 75.

[0028]

[Problems to be Solved by the Invention] However, in the above configuration, all orthogonal transform coefficients smaller than the quantization step size Qb become "0" after quantization, so (1) the quantization step size Qb is (2) If there are many orthogonal transform coefficients with a small difference between the orthogonal transform coefficients and the quantization step size, "block distortion" and "mosquito noise", which are deterioration in the quality of the reproduced image due to quantization, will occur. There was a problem.

The reason for this is that the orthogonal transform coefficient has a visually important meaning, and its value is the quantization step size Qb.
If it is even slightly smaller than , the orthogonal transform coefficient after quantization becomes "0", and the quality of the reproduced image is considered to have deteriorated. In particular, the quantization step size Q
When b is large, many orthogonal transform coefficients after quantization are
0'', the quality of the reproduced image deteriorates significantly.

In view of the above problems, the present invention calculates the absolute value |C of each orthogonal transform coefficient before quantizing the orthogonal transform coefficient.
oef | and the quantization step size Qb, and among the orthogonal transform coefficients that become "0" after quantization, for the orthogonal transform coefficients for which the difference between the orthogonal transform coefficient and the quantization step size is equal to or less than a predetermined threshold Th, Absolute value of orthogonal transformation coefficient |C
By performing numerical conversion using oef| as the quantization step size, it is possible to prevent orthogonal transform coefficients whose difference in quantization step size is less than or equal to a predetermined threshold Th after quantization from becoming "0", and to The purpose is to improve the image quality of reproduced images by preventing an increase in the amount of code that occurs due to an increase in the number of events that occur due to the conversion coefficient being subjected to the numerical conversion process.

[0031]

Means for Solving the Problems In order to achieve the above object, the technical solution of the present invention provides an analog-to-digital conversion means for converting a television signal from an analog signal to a digital signal, and a digital input television signal. a blocking means for dividing one frame or one field into blocks of a predetermined size; a motion vector calculating means for calculating a motion vector representing the motion of a television image for each block; a motion compensation determination means for determining whether to perform motion compensation prediction using the vector;
A motion compensation predicted pixel calculation means for calculating a predicted pixel value, a prediction error calculation means for calculating the difference between the pixel value of the input television signal and the predicted pixel value as a prediction error value, and interframe coding for each block. encoding method determining means for determining whether to perform intra-frame encoding or intra-frame encoding; and whether a signal to be orthogonally transformed based on the intra-frame encoding/inter-frame encoding determination result for each block is used as a pixel value of an input television signal. orthogonal transform signal selection means for converting or switching a prediction error value; orthogonal transform means for orthogonally transforming pixel values or prediction error values of an input television signal to calculate orthogonal transform coefficients; The quantization step size calculation means to be calculated compares each orthogonal transform coefficient with the quantization step size, and calculates the corresponding orthogonal transform for orthogonal transform coefficients whose orthogonal transform coefficient value becomes zero after quantization within a predetermined range. orthogonal transform coefficient converting means for numerically converting coefficient values to a quantization step size; and a first orthogonal transform that quantizes all orthogonal transform coefficients including the numerically converted orthogonal transform coefficients using the quantization step size. a coefficient quantization means, a first event code amount calculation means that encodes the numerically converted and quantized orthogonal transform coefficients in units of generated events and calculates the generated code amount, and an orthogonal transform coefficient that does not undergo the numerical conversion process. a second orthogonal transform coefficient quantizer that quantizes the quantized orthogonal transform coefficient using a quantization step size; and a second event that encodes the quantized orthogonal transform coefficient in units of generated events and calculates the amount of generated code. The code amount calculation means compares the generated code amount of an event without numerical conversion with the total generated code amount of multiple events that occur within the range of events when numerical conversion is not performed, and calculates the generated code amount. means for calculating generated code amount information for each event, which calculates the smaller of the
means for selecting orthogonal transform coefficients with a small amount of generated codes from the output of the first orthogonal transform coefficient quantization means and the output of the second orthogonal transform coefficient quantization means, using the generated code amount information in units of events; An encoding means for encoding information on intra-frame encoding or inter-frame encoding, a motion vector value, a quantization step size, and a quantized orthogonal transform coefficient; An inverse orthogonal transform means for calculating a quantized signal and a pixel value to be used when calculating a reproduced pixel value based on the interframe coding/intraframe coding determination result, whether the pixel value is a predicted pixel value subjected to motion compensation prediction, or a numerical value "0"
A motion compensation device comprising a reproduction pixel selection means for selecting or switching the reproduction pixel, a reproduction pixel calculation means for calculating a reproduction image from a predicted pixel value or numerical value "0" and an inverse quantization signal, and a reproduction pixel accumulation means for accumulating the reproduction image. The above object is achieved by a predictive interframe coding device.

[0032]

[Operation] When measuring the distribution of the orthogonal transform coefficient Coef subjected to quantization processing, whether the orthogonally transformed image is an intra-frame image or an inter-frame image, (1) the absolute value of the coefficient |Coef in the low frequency component region There are many distributions where | is large, and (
2) Many coefficients with small absolute values |Coef| are distributed in the high frequency component region. Therefore, it is often the orthogonal transform coefficient values belonging to the high frequency component region that are converted to the numerical value "0" after quantization.

If the high frequency component of the orthogonal transform coefficient of the intra-frame image becomes a numerical value "0" due to quantization, the reproduced image will be a blurred image.

[0034] If the high-frequency component of the orthogonal transform coefficient of the interframe image becomes a numerical value "0" due to quantization, the reproduced image will contain "
Image quality deterioration such as "mosquito noise" occurs.

On the other hand, the orthogonal transform coefficients after quantization are shown in Figure 3 (
Using a zigzag scan as shown in a), divide into a continuous number of coefficient values ``0'' (run length) and a pair of coefficient values that are not ``0'' (this is called an ``event''),
The code length is calculated for each "event" using the variable length code code length table shown in FIG. 3(b). In a variable length code, a code with a short code length is assigned to an event that occurs frequently, and a code with a long code length is assigned to a code that occurs less frequently. Therefore, when numerical conversion processing is performed on the orthogonal transform coefficients, one event is divided into two or more events, so the total amount of code generated for two or more events increases, and the quality of the reproduced image may deteriorate. Therefore, the present invention solves the image quality deterioration that is thought to occur due to quantization by using the above configuration, before quantizing the orthogonal transform coefficients, each orthogonal transform coefficient is The absolute value of the coefficient |Coef| is compared with the quantization step size Qb, and for orthogonal transform coefficients for which the difference between the two is less than or equal to a predetermined threshold Th, the absolute value of the orthogonal transform coefficient |C
By performing numerical conversion using oef| as the quantization step size, it is possible to prevent orthogonal transform coefficients having a size larger than a certain level from becoming "0" after quantization, and furthermore, the numerical conversion process of orthogonal transform coefficients can be Compare the total generated code amount of 2 or more events that occurred as a result of this process with the generated code amount of one event that occurs without performing numerical conversion processing of orthogonal transform coefficients, and calculate the generated code amount by numerical conversion processing of orthogonal transform coefficients. By selecting whether or not to perform numerical conversion processing on the orthogonal transform coefficients, the amount of generated code is prevented from increasing, and as a result, the quality of the reproduced image is improved.

[0036].

Embodiments Hereinafter, embodiments of the present invention will be described with reference to FIG. FIG. 1 is a block diagram of a motion compensated predictive interframe coding apparatus in an embodiment of the present invention. In FIG. 1, 1 is an input terminal into which an input television signal is input, and 3 is a motion vector of the coded block and a motion compensation prediction control signal by comparing the picture signal of the coded block of the current frame and the reproduced picture signal of the previous frame. 4 is an image memory unit that stores the reproduced image signals of the current frame and the previous frame; 7 is a motion compensation prediction unit that performs motion compensation prediction on the reproduced image signal of the previous frame; 9 is an encoding unit; 11 is an inter-frame/intra-frame determining unit that determines whether a block to be encoded is to be intra-frame encoded or intra-frame encoded; 11 is an intra-loop filter unit that performs two-dimensional low-pass filter processing on the motion-compensated prediction signal; 13 is an encoding unit; A prediction error calculation unit calculates a prediction error by calculating the difference between the original image signal and the prediction signal of the block, and 15 is a switch that selects a signal to be orthogonally transformed and a signal for calculating a reproduced image based on an encoding method selection signal. 17 is an orthogonal transformation unit that performs orthogonal transformation; 19 is an orthogonal transformation coefficient transformation unit that numerically transforms orthogonal transformation coefficients; 20 is a quantization size calculation unit that calculates a quantization step size; and 23 is an orthogonal transformation coefficient that has been numerically transformed. 25 and 29 are event encoding units that calculate code lengths for each event of orthogonal transform coefficients;
27 is an orthogonal transform coefficient quantization unit that quantizes orthogonal transform coefficients that are not numerically converted; 31 is a generated code amount comparison unit that determines whether or not to select numerically converted orthogonal transform coefficients from the generated code amount for each event; 33 3 is a switch unit that selects orthogonal transform quantization coefficients for each event based on the selection signal determined by the generated code amount comparison unit; 35 is a code memory unit that stores information on the transmission frame format; and 37 is a switch unit that inverts orthogonal transform quantization coefficients. an inverse orthogonal transform unit that performs orthogonal transformation; 41 is a reproduced image calculation unit that calculates a reproduced image of the current frame; 43 is a prediction error encoding unit that encodes the encoding method selection signal, prediction error, and quantization step size through channel coding; 45 is a motion vector encoding unit that encodes a motion vector through a channel; 47 is a multiplexer unit that configures a transmission frame from a prediction code and a motion vector code; and 50 is an output terminal that outputs a transmission signal.

The operation of the above configuration will be explained below. The television signal is converted from analog to digital by a signal processing section not shown in FIG. Next, the motion vector calculation section 3 compares the input television signal 2 with the reproduced image 5 of the previous frame read out from the image memory section 4, calculates a motion vector, and outputs it as a motion vector signal 6. At the same time, the motion vector calculation unit 3 uses the evaluation value at the time of motion vector calculation to determine whether motion compensation prediction for the encoded block is valid or invalid, and outputs the result as motion compensation prediction control information to the motion vector signal 6. do.

The motion compensation prediction unit 7 performs motion compensation prediction using the motion vector when performing motion compensation prediction using the motion vector signal 6 on the reproduced image 5 of the previous frame at the same position as the encoded block, and when not performing motion compensation prediction. The reproduced image of the previous frame is output as is as a motion compensated prediction signal 8.

The inter-frame/intra-frame determining unit 9 compares the input television signal 2 and the motion compensated prediction signal 8 on a block-by-block basis, determines the effectiveness of inter-frame prediction, and determines if the effectiveness of inter-frame prediction is small. determines that intraframe coding is effective for the corresponding block, and if the effectiveness of interframe prediction is large, determines that interframe coding is effective for the corresponding block, and sends the result to coding method selection signal 1.
Output as 0.

The in-loop filter unit 11 performs in-loop filter processing, which is two-dimensional low-pass filter processing, on the motion-compensated prediction signal 8 if the encoded block is a block to be predicted with motion compensation, and in other cases, performs in-loop filter processing on the motion-compensated prediction signal 8. The predicted signal 12 is output without being filtered.

The prediction error calculation section 13 calculates the difference between the input television signal 2 and the prediction signal 12 of the encoded block, and outputs the result as a prediction error signal 14.

The switch section 15 (1) selects the input television signal 2 as the signal 16 to be orthogonally transformed when the encoding method selection signal 10 selects intraframe encoding, and (2) selects the input television signal 2 as the signal 16 to be orthogonally transformed. When the method selection signal 10 selects interframe coding, the prediction error signal 14 is selected as the signal 16 to be orthogonally transformed.

The orthogonal transform unit 17 converts the signal 16 to be orthogonally transformed.
The high correlation between neighboring pixels of the signal 16 to be orthogonally transformed is removed, and the orthogonal transform coefficient 1 is
Calculate 8. As the orthogonal transform method, in many cases, discrete cosine transform is used, which has high transform efficiency and can be implemented in hardware.

The orthogonal transform coefficient converting unit 19 compares the absolute value |Coef| of the orthogonal transform coefficient 18 with the quantization step size 21 calculated by the quantization step size calculating unit 20, and calculates the difference between the two from a predetermined threshold Th. If it is not large, numerical conversion is performed using the absolute value |Coef| of the orthogonal transformation coefficient as the quantization step size Qb, and the numerically transformed orthogonal transformation coefficient 22 is calculated. In other words.

The first quantizer 23 quantizes the numerically converted orthogonal transform coefficients 22 using the quantization step size 21 to calculate orthogonal transform quantized coefficients 24 .

[0046] The first event encoding unit 25 converts the orthogonal transform coefficients 22 in units of blocks into

[0047]

[Table 1]

Scanning is performed in accordance with the zigzag scanning order shown in [0048] and the orthogonal transformation coefficients are converted into events. In addition,(
In Table 1), the scanning order of the orthogonal transform coefficients (8*8) is shown, and it is assumed that scanning starts at "1" and ends at "64". Next, the event encoding unit 25 determines the code length of each event.

[0049]

[Table 2]

Using the code length table of the variable length code shown in
It is calculated as the generated code amount 26. Note that (Table 2) shows the code length of the two-dimensional Huffman code of the event, and blank spaces are omitted.

Similarly, the second quantizer 27 uses the quantization step size 21 to quantize the orthogonal transform coefficients 18 that have not been numerically transformed, and calculates the orthogonal transform quantized coefficients 28.

The second event encoding unit 29 scans the orthogonal transform coefficients 28 block by block in accordance with the zigzag scan order shown in Table 1, and converts the orthogonal transform coefficients into events. Next, the event encoding unit 29 calculates the code length of each event as the generated code amount 30 using the variable length code code length table shown in (Table 2).

The code amount comparison unit 31 compares the generated code amount 30 of an event in which the orthogonal transform coefficients are not subjected to numerical conversion processing with the generated code amount 26 of an event in which the orthogonal transform coefficients are subjected to numerical conversion processing, and determines the generated code. Select low volume events,
An event selection signal 32 is output.

Next,

[0055]

[Table 3]

[0056]

[Table 4]

An example of event selection is shown below. Note that (Table 3) shows orthogonal transformation coefficients that are not numerically converted, and (Table 4) shows orthogonal transformation coefficients that are numerically converted. event, (run length of coefficient "0", coefficient value of coefficient not "0")
In the case shown in (Table 3), some of the events of the orthogonal transformation coefficients shown in (Table 3) are as follows.

(3, 1), (1, 2), (0, 2), (
6, 3),...If the orthogonal transform coefficients shown in (Table 3) are numerically converted and the orthogonal transform coefficients become as shown in (Table 4), the event is as follows. It becomes like this.

(3, 1), (1, 2), (0, 2), (
2, 1), (3, 3), etc. Therefore, the event (6, 3) becomes the events (2, 1), (3, 3) by the numerical transformation of the orthogonal transformation coefficients. Therefore, the event (j
, k), and m(j, k) is the generated code amount of (
From the code length table shown in Table 2), the code length is as follows.

m(6,3)=20bitm(2,1)=
5 bits m (3, 3) = 13 bits Therefore, m (6, 3) > m (2, 1) + m (3, 3), so select the numerically converted event.

The switch section 33 receives the event selection signal 32
Select the event signal by. In the case of the above example, events (2, 1) and (3, 3) are selected and output as orthogonal transform quantization coefficients 34.

The quantization step size 21 is calculated by the quantization step size calculation section 20 from the code residual amount 36 in the code memory section 35 using the method described in the conventional example.

The inverse orthogonal transform unit 37 performs inverse orthogonal transform on the orthogonal transform quantization coefficients 34 to calculate a signal 38 containing a quantization error.

The switch section 15 (1) selects the numerical value "0" signal 40 as the reproduced image calculation signal 39 when the encoding method selection signal 10 selects intraframe encoding, and (2) selects the numerical value "0" signal 40 as the reproduction image calculation signal 39. When the coding method selection signal 10 selects interframe coding, the prediction signal 12 is selected as the reproduced image calculation signal 39.

The reproduced image calculation unit 41 adds the signal 38 containing the quantization error and the reproduced image calculation signal 39 to calculate a reproduced image 42 of the encoded block.

The image memory 4 stores the reproduced image signal 42 of the current frame and outputs the reproduced image signal 5 of the previous frame.

The prediction error encoding unit 43 encodes the encoding method selection signal 10, the quantization step size 21, and the orthogonal transform quantization coefficient 34, and calculates the prediction error code 44.

The motion vector encoding unit 45 encodes the motion vector signal 6 of the block subjected to motion compensation prediction and calculates a motion vector code 46.

The multiplexer unit 47 uses the prediction error code 44
A transmission frame 48 in a predetermined format is calculated from the motion vector code 46.

The code memory unit 35 temporarily stores the transmission frame 48 and outputs it from the output terminal 50 as a transmission code 49 in synchronization with a clock signal input from an external device (not shown). At the same time, the code memory unit 35 calculates the amount of codes remaining in the memory as the remaining code amount 36.

As is clear from the above description, according to this embodiment, when the orthogonal transform coefficients are quantized, the orthogonal transform coefficients that were quantized to a numerical value of "0" because they were slightly smaller than the quantization step size are Numerical conversion processing of the orthogonal transform coefficients remaining after quantization, total amount of generated codes of 2 or more events that occur as a result of performing the numerical conversion processing of the orthogonal transform coefficients, and 1 that occurs without performing the numerical conversion processing of the orthogonal transform coefficients. By comparing the generated code amount of events and selecting whether or not to perform the numerical conversion process of the orthogonal transform coefficient, the increase in the generated code amount can be prevented so that the generated code amount does not increase due to the numerical conversion process of the orthogonal transform coefficient. As a result, the quality of reproduced images is improved.

[0072]

As described above, in the present invention, before quantizing orthogonal transform coefficients, for orthogonal transform coefficients that become "0" after quantization, the absolute value of each orthogonal transform coefficient |Coef| By comparing the step size Qb and performing numerical conversion using the absolute value |Coef| of the orthogonal transform coefficient as the quantization step size for orthogonal transform coefficients whose difference between the two is less than or equal to a predetermined threshold Th, it is possible to 2, which occurred as a result of numerical conversion processing of orthogonal transform coefficients, to prevent orthogonal transform coefficients with a size larger than 0 from becoming "0".
Compare the total amount of codes generated for an event or more with the amount of codes generated for one event that occurs without performing numerical conversion processing of orthogonal transform coefficients, and calculate the orthogonal By selecting whether or not to perform numerical conversion processing of the conversion coefficients, the image quality of the reproduced image can be improved by preventing an increase in the amount of generated codes, which has a large effect.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of a motion compensated predictive interframe coding device in an embodiment of the present invention.

[Fig. 2] Block wiring diagram of a conventional motion compensated predictive interframe coding device

[Explanation of symbols]

1, 51 Input terminals 3, 53 Motion vector calculation units 4, 54 Image memory units 7, 58 Motion compensation prediction units 9, 60 Intra-frame/inter-frame determination units 11, 62
In-loop filter section 13, 64 Prediction error calculation section 15, 33, 66 Switch section 17, 68 Orthogonal transformation section 19 Orthogonal transformation coefficient transformation section 20, 73 Quantization step size calculation section 23, 27
, 70 quantization units 25, 29 event encoding unit 31 code amount comparison unit 35, 74 code memory unit 37, 76 inverse orthogonal transformation unit 38, 78 reproduced image calculation unit 43, 82 prediction error encoding unit 45, 84 motion vector Encoding section 47, 86 Multiplexer section 50, 89 Output terminal

Claims (1)

[Claims] 1. Analog-to-digital conversion means for converting a television signal from an analog signal to a digital signal, and blocking means for dividing one frame or one field of a digitized input television signal into blocks of a predetermined size. a motion vector calculation means for calculating a motion vector representing a motion of a television image for each block; a motion compensation determination means for determining whether or not to perform motion compensation prediction using the motion vector for each block; For a block to be predicted with motion compensation, a motion compensation prediction pixel calculation means that performs motion compensation prediction on the reproduced image of the previous frame using a motion vector and calculates a predicted pixel value, and a motion compensation prediction pixel calculation means that calculates a predicted pixel value, and a difference between the pixel value of the input television signal and the predicted pixel value. A prediction error calculation means for calculating a prediction error value; a coding method determination means for determining whether to perform interframe encoding or intraframe encoding for each block; orthogonal transform signal selection means for switching a signal to be orthogonally transformed as a pixel value of an input television signal or a prediction error value according to an encoding determination result; an orthogonal transform means for calculating an orthogonal transform coefficient; a quantization step size calculation means for calculating a quantization step size from the generated code amount; For orthogonal transform coefficients that become 0 after quantization within a predetermined range, an orthogonal transform coefficient converting means for numerically converting the corresponding orthogonal transform coefficient value to a quantization step size, and an orthogonal transform coefficient subjected to numerical conversion processing are included. a first orthogonal transform coefficient quantizer that quantizes all orthogonal transform coefficients using a quantization step size; and a first orthogonal transform coefficient quantizer that quantizes all orthogonal transform coefficients using a quantization step size; a first event code amount calculation means that pairs the number of events and the value of a coefficient whose value is not 0, encodes the events of this pair in event units, and calculates the generated code amount; and orthogonal transformation that does not perform the numerical conversion process. a second orthogonal transform coefficient quantizer that quantizes coefficients using a quantization step size; and a second event code that encodes the quantized orthogonal transform coefficients in units of events and calculates the amount of generated codes. The amount calculation means compares the amount of code generated by an event without numerical conversion, and the total amount of code generated for multiple events that occur within the range of the event without converting the numerical value, and calculates the generated code. The output of the first orthogonal transform coefficient quantization means and the second orthogonal transform are calculated by using the event-based generated code amount information calculation means that calculates the smaller amount, and the event-based generated code amount information. Selection means for selecting an orthogonal transform coefficient with a small amount of generated code from among the outputs of the coefficient quantization means, information on intra-frame coding or inter-frame coding, motion vector value, quantization step size, and quantized orthogonal An encoding means for encoding the transform coefficients, an inverse orthogonal transform means for performing inverse orthogonal transform on the quantized orthogonal transform coefficients and calculating an inverse quantized signal, and reproduction based on the interframe encoding/intraframe encoding determination result. Either set the pixel value used when calculating the pixel value as the predicted pixel value obtained by motion compensation prediction, or use the numerical value "0".
”; a reproduction pixel calculation means for calculating a reproduction image from a predicted pixel value or numerical value “0” and an inverse quantization signal; and reproduction pixel accumulation means for accumulating the reproduction image. Compensated predictive interframe coding device.