JPH04307887A - Transmission rate control system for picture encoding - Google Patents

Abstract

PURPOSE:To efficiently control the transmission rate without increasing the code volume by sending the code of coefficients till the transmission code volume exceeds a set value and forcibly stopping to send the code for remaining coefficients. CONSTITUTION:The code volume on the way of scanning is successively calculated in a rate calculating circuit 108 and is inputted to a variable length code encoding control circuit 109 as a momentary rate Rp and an average rate Rh. Rate set values rh (average) and rp (peak) are preliminarily given to the circuit 109, and a control signal Cp or Ch is sent if the average value or the peak value exceeds them. In the case of Rp>rp, a variable length code encoding part 107 regards remaining coefficients as 0 to terminate encoding of the block. In the case of Rh>rh, an EOB code is forcibly inserted after a two-dimensional code to abandon the remaining code, and encoding of the block is terminated. Thus, an influence of the visual degradation in picture quality is less because abandoned coefficients are on the high frequency component side.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a transmission rate control function in interframe and intraframe encoding of images.

0002

2. Description of the Related Art The following documents describe an image intraframe encoding device having a transmission rate control function. Literature title: "Characteristics of a variable rate image encoder with average rate control", 1st Information Transmission and Signal Processing Workshop, March 16 and 17, 1989 Figure 2 shows the characteristics of a conventional encoder. A configuration diagram is shown. In FIG. 2, an A/D converted and digitized image signal Sij is input to an input terminal 201.
One frame is divided into blocks each consisting of 8×8 pixels, and each block is applied to the subtraction circuit 204 and the motion compensation circuit 202. Furthermore, as will be described later, the frame memory 203 stores an image signal of one frame before the current frame, and this signal of one frame before is also given to the motion compensation circuit 202. The motion compensation circuit 202 detects the amount of motion by comparing the image signals of the current frame and the previous frame in block units, and forms a prediction signal SPij of the image signal of the current frame from the previous frame according to the detected amount of motion. and is input to the subtraction circuit 204.

SPij obtained in this way and the input signal Sij are subtracted by a subtraction circuit 204, which calculates the difference (S-SP).
is taken and is indicated by the difference signal Eij. Difference signal Ei
j is input to the DCT calculation circuit 205, subjected to DCT calculation, and quantized by the quantizer 206. The quantized value Iij obtained by quantization is input to the variable length encoder 207 and the inverse quantizer 210. Here, for the quantizer 206 and the inverse quantizer 210, several types with different quantization step sizes are prepared in addition to the quantizer and inverse quantizer provided initially. The variable length encoding unit 207 performs zigzag scanning for each 8×8 pixel block, which is a processing unit of DCT coefficients, until the last non-zero coefficient.
It is expressed as a two-dimensional code (run length, level) by a pair of the magnitude (level) of a non-zero coefficient and the number of zeros (run length) sandwiched between non-zero coefficients. After scanning up to the last non-zero coefficient, the remaining zero coefficients are discontinued and EOB (
(end of block) code is attached as a code. FIG. 3 shows an example of variable length encoding using two-dimensional encoding. The block shown in FIG. 3 is {(0,5), (1,
3), (7, 1), EOB}, and the code for this is assigned a Huffman code according to the frequency of occurrence through training.

[0004] The code obtained in this way is sent out to the transmission path, and at the same time is input to the rate calculation circuit 208.
The average amount of codes generated within a certain period of time (for example, one frame or one second) and the instantaneous peak amount of codes (for example, for each block) are calculated and transmitted to the quantizer control circuit 209. Rate setting values Rh (average) and Rp (peak) are given in advance to the quantizer control circuit 209, and when either the average code amount or the peak code amount exceeds these, the quantizer 206 and the reverse A quantization control signal c is sent to the quantizer 207, and the quantization step size is switched to one with a larger quantization step size. This increases the number of invalid pixels of DCT coefficients and suppresses the amount of generated code. Further, when the average code amount and the peak code amount fall within the set values, the quantizer and inverse quantizer are returned to their initial values. This is based on the fact that DCT coefficients generally have low power on the high frequency side, so that 0 occurs more frequently on the high frequency side. Furthermore, how many coefficients are encoded and transmitted depends on the quantization characteristics of the quantizer (quantization step size, number of quantization bits, etc.). On the other hand, inverse quantizer 2
After being dequantized with the quantization value Iij input to 10,
An inverse DCT operation is performed in an inverse DCT circuit 211, and an adder 211
By adding the predicted signal SPij with the predicted signal SPij, a locally reproduced signal Sd at the encoder is obtained. Furthermore, the local reproduction signal S
d is stored in the frame memory 203 and used to generate the prediction signal SPij for the next frame.

[0005]

[Problems to be Solved by the Invention] In the conventional transmission rate control system, the amount of code generated from the encoder is monitored, and when the amount of code generated from the encoder exceeds a preset value, the quantization step size of the quantizer is increased to reduce the amount of code generated. This is to reduce the amount. In this case, it is necessary to send to the receiving side information indicating that the quantization step size has been switched, and additional information indicating whether the quantization step size has been switched for a block of the frame. This causes a problem in that the amount of information to be transmitted increases. Furthermore, when the quantization step of the low frequency component of the block is switched to a large one, the influence on the image quality is large. Therefore, an object of the present invention is to control the transmission rate during interframe encoding of images without sending special additional information to the receiving side.

[0006]

[Means for Solving the Problems] The present invention performs zigzag scanning of quantized values of transform coefficients of a discrete cosine transform, and provides a two-dimensional method using pairs of the number of 0s between non-zero coefficients and the size of the non-zero coefficients. Perform variable length encoding using coding,
At this time, the code amount is calculated at the same time, and when it exceeds a certain set value, the scan is terminated midway, and the remaining coefficients are set to 0 (invalid).

[0007]

[Operation] In this transmission rate control method, when the transmission code amount exceeds the set value, the codes of the coefficients up to the middle are sent, and the codes for the remaining coefficients are not forced to be sent. The number of codes to be used can be reduced. Furthermore, since there is no change in the characteristics of the quantizer, the receiving side only needs to decode the transmitted information, and there is no need to particularly send additional information to the receiving side.

[0008]

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 shows a block diagram of an encoding apparatus according to a first embodiment of the present invention. An A/D converted and digitized image signal S is input from an input terminal 101, one frame is divided into blocks each consisting of 8×8 pixels, and each block is provided to a subtraction circuit 104 and a motion compensation circuit 102. . In addition, the frame memory 103 stores an image signal of one frame before the current frame, as will be described later, and this signal of one frame before is also stored in the motion compensation circuit 103.
Given to 02. The motion compensation circuit 102 detects the amount of motion by comparing the image signals of the current frame and the previous frame block by block, and forms a prediction signal SPij of the image signal of the current frame from the previous frame according to the detected amount of motion. and input to the subtraction circuit 104. The difference (S-SP) between SPij obtained in this way and the input signal Sij is taken by a subtraction circuit 104, and this is indicated as a difference signal Eij. The difference signal Eij is input to the DCT calculation circuit 105, subjected to DCT calculation, and quantized by the quantizer 106. The quantized value Iij obtained by quantization is input to the variable length encoding section 107.

In the variable length encoding unit 107, if control signals Rp and Rh, which will be described later, are not provided, zigzag scanning is performed for each 8×8 pixel block, which is a processing unit of DCT coefficients, until the last non-zero coefficient. , expressed as a two-dimensional code (run length, level) by a pair of the magnitude (level) of a non-zero coefficient and the number of zeros (run length) sandwiched between non-zero coefficients. Then, after scanning until the last non-zero coefficient,
The remaining 0 coefficients are truncated and EOB (end of
A block) code is added to make it a transmission code, and it is sent to the transmission path, and at the same time it is sent to the receiving side, it is processed by the variable length decoder 11.
It is input to 0.

During two-dimensional encoding in variable length encoding section 107, the rate calculation circuit 108 sequentially calculates the amount of codes during scanning, and transmits the result to variable length encoding control circuit 109 as an instantaneous rate Rp. Further, the code sent from the variable length encoding section 107 is sent to the receiving side and simultaneously inputted to the rate calculation circuit 108, and is sent to the variable length encoding control circuit 109 as the average rate Rh per unit time (for example, per frame). is input. Variable length encoding control circuit 1
09, the rate setting values rh (average), rp
(peak) is given, and when either the average code amount or the peak code amount exceeds this, the control signal cp
, ch are sent, and the variable length encoder 107 is controlled as follows.

i) When Rp>rp, during the zigzag scan, when the control signal cp is received, an EOB code indicating the end of encoding is forcibly inserted, and the remaining coefficients are set to 0, and the encoding for that block is performed. end. ii) When a control signal ch is received when Rh>rh, an EOB code is forcibly inserted after m two-dimensional codes (run length, level pairs) set in advance, and the remaining codes are discarded. Then, encoding for that block ends. For example, the results of zigzag scan are (0,8), (3,6), (4,4), (1
,5),(6,2),(8,1),(3,1),EOB
When a two-dimensional code such as (0,8), (3,
6), (4, 4), (1, 5), (6, 2), transmitted to the receiving side as EOB. However, this is an example when m=5. By suppressing the amount of code to be transmitted in this way, the transmission rate is controlled. Since the coefficients that are discarded by rate control are on the high frequency component side, the effect of visually deteriorating image quality is less than when the coefficients on the low frequency side are lost.

On the other hand, the code input to the variable length decoder 110 is variable length decoded to obtain a quantized value Id. Furthermore, the quantized value Id is inversely quantized by an inverse quantizer 111, then subjected to an inverse DCT operation by an inverse DCT circuit 112, and added to the prediction signal SPij by an adder 113, whereby local reproduction in an encoder is performed. Obtain signal Sd. Furthermore, the local reproduction signal Sd
is stored in the frame memory 103 and used to generate the prediction signal SPij for the next frame.

FIG. 4 shows a configuration diagram when the present invention is applied to intraframe coding, which is a second embodiment. The A/D converted and digitized image signal s is input to the input terminal 114.
One frame is divided into blocks each consisting of 8×8 pixels, and each block is processed by a DCT calculation circuit 114.
and is quantized by a quantizer 115. Each transform coefficient quantized by the quantizer 115 is processed by the variable length encoder 11
In step 6, two-dimensional run-length encoding using zigzag scanning is performed, and the signal is sent to the transmission path. At this time, the rate calculation circuit 117 calculates the average code amount and the instantaneous code amount per unit time, as in the first embodiment, and the variable length coding control circuit 118 controls the variable length coding section. The transmission rate is controlled by this.

[0014]

[Effects of the Invention] In the transmission rate control method for an image code having the above configuration, zigzag scanning is performed and two-dimensional encoding is performed using a pair of the number of 0s between non-zero coefficients and the size of the non-zero coefficients. The number of codes to be transmitted can be reduced by using variable length coding to send codes for the coefficients up to the middle of the block and not forcibly sending codes for the remaining coefficients. Furthermore, since there are no changes to the characteristics of the quantizer, the receiving side only needs to decode the information sent, and there is no need to send any additional information to the receiving side, so there is no increase in the amount of code. , efficient transmission rate control becomes possible. Since the truncated coefficients are also cut from the high-frequency components within the block, the visual impact is less than the deterioration of the low-frequency coefficients.

[Brief explanation of drawings]

FIG. 1 is a configuration diagram of an encoder according to a first embodiment of the present invention.

[Figure 2
]Explanatory diagram of conventional technology

[Figure 3] Explanatory diagram of variable length encoding using two-dimensional encoding

[
FIG. 4: Configuration diagram of an encoder according to a second embodiment of the present invention

[Explanation of symbols]

102 Motion compensation section 103 Frame memory 104 Subtractor 105 DCT operation section 106 Quantizer 107 Variable length encoder 110 Variable length decoding unit 111 Inverse quantizer 112 Inverse DCT operation section 113 Adder

Claims (2)

[Claims] Claim 1: Discrete cosine transform is applied to each pixel block, and zigzag scanning is performed for each block of the quantized value of each coefficient value, and the number of 0s between non-zero coefficients and the size of the non-zero coefficients are calculated. In an image coding method that performs variable length coding using two-dimensional coding using pairs of 3D and 3D, the average code amount of transmission codes is calculated, and it is determined whether this average code amount exceeds a preset code amount. When the average code amount exceeds the preset code amount, a control signal is generated, and while the control signal is being generated, during the variable length encoding, for each block, A transmission rate control method for image encoding, characterized in that a preset number of pairs of 0's and coefficient sizes are encoded, and the remaining coefficients are set to 0. 2. In the transmission rate control method for image encoding according to claim 1, the instantaneous code amount of the transmission code is calculated, it is detected whether this instantaneous code amount exceeds the code amount separately set in advance, and the When the instantaneous code amount exceeds the code amount set separately in advance, another control signal is generated, and at the time this control signal is generated, the variable length encoding of the block is forcibly stopped. Features a transmission rate control method in image encoding.