JPH03229586A - Processing quantity adaptive type motion compensating moving picture encoding system - Google Patents

Abstract

PURPOSE:To bring out the processing capacity of hardware to the maximum without lowering encoding efficiency by changing the number of searching blocks for motion compensation according to the number of transmitting frames per unit time. CONSTITUTION:A motion compensation detection circuit 23 is provided with a function to change the number of the searching blocks for motion compensation according to the number of the transmitting frames per unit time when it obtains an optimum vector. An evaluation calculation circuit 232 calculates the block ji in which the evaluation value Si of the block ji becomes minimum from data dk in the block (i) of an original picture and the data dk' in the block ji of a predicted picture, and reads out the corresponding optimum vector from a vector list 231, and outputs it to a variable delaying part 22 and a receiving part 5. Thus, since the number of the searching blocks is changed according to the number of the transmitting frames per unit time, the capacity of the hardware can be exhibited to the maximum, and in addition, the encoding efficiency is never greatly lowered.

Description

[Detailed Description of the Invention] [Summary] By dividing the screen into blocks of a certain size regarding image information and searching for the block with the minimum prediction error from the prediction screen for each block, high Regarding the motion compensated video coding method that performs efficiency coding, the purpose is to be able to adaptively switch the processing amount according to the number of transmitted frames per unit time. The number of search blocks for motion compensation is configured to be changed depending on the motion compensation.

[Industrial Application Field] The present invention divides the screen into blocks of a certain size for one image information, and searches for a block with the minimum prediction error from the prediction screen for each of these blocks. The present invention relates to a motion compensated video encoding method that performs efficiency encoding.

For example, when it comes to image signals from video calls and video conferences, the corresponding images between two frames generally have similar values, so the information between such frames has a strong correlation. For this purpose, the image signal is subjected to band compression encoding, and at this time, a prediction error is obtained by applying motion compensation to the moving image.

Here, the motion compensation method is one of the video band compression methods, and is also called the motion compensation interframe coding method, which utilizes the correlation between frames and detects motion. This method greatly reduces redundancy.

[Prior Art] FIG. 6 is a system configuration diagram for explaining the motion compensation method. In FIG. , the motion compensation means 102 performs motion compensation by calculating a prediction error from the least squares error, absolute value error, etc., the quantization means 103 performs quantization, and the encoding means 104 encodes the input signal. Receiver part 1
05 is performed.

Then, in the receiving section 105, the received signal is decoded by the decoding means 10.
6, the motion compensation means 107 performs motion compensation in the same manner as the transmission section 101, and the dequantization means 108 dequantizes and outputs it as a reproduced image.

[Problems to be Solved by the Invention] In the motion compensated video encoding method, a screen is divided into blocks of a certain size (for example, 8 pixels x 8 lines), and prediction errors are calculated from the predicted screen for each block. A search for the smallest block is performed. Therefore, when the number of blocks to be searched increases,
The processing amount also increases in proportion to this.

Now, if we consider a normal NTSC signal as the image signal to be compressed, 30 frames are input per second.

Generally, it is desirable to transmit all 30 screens per second, but depending on the input scene, the number of transmitted screens varies greatly, and in scenes with large movements or scene changes, predictions may not be accurate, and in this case, 1 Since the amount of information generated per screen increases, the number of transmitted screens is significantly reduced, for example to 5-10 frames/sec.

In consideration of such cases, in the conventional design method, the circuit for performing the search processing necessary for the above-mentioned motion compensation prediction is designed to have enough hardware to transmit 30 screens in any case. As a result, the circuit scale becomes large, and when the number of transmitted screens is reduced due to scenes with large movements, the problem is that the processing power of the hardware is wasted.

By the way, generally speaking, if you increase the number of search blocks for motion compensation, it will be possible to track large movements, reduce the power of the prediction error signal, and increase coding efficiency, but this will require more processing power. .

Usually, the processing power of hardware is fixed, so it is necessary to use it to the maximum extent so as not to reduce encoding efficiency.

Generally, in a scene with large movements, it is necessary to search for as many blocks as possible, but at this time, the amount of information generated per screen is large, so the number of transmitted frames decreases. Furthermore, in a static scene with little movement, it is not necessary to search a large number of blocks, and the amount of information generated per screen at this time is small, so the number of transmitted screens can be increased.

On the other hand, when the number of transmitted frames decreases, hardware with a certain processing power can search many blocks,
Conversely, as the number of transmitted frames increases, the number of blocks that can be searched decreases.

Therefore, when the number of transmitted frames drops due to large movements,
By increasing the number of search blocks and decreasing the number of search blocks when the motion is small and the number of transmission frames increases, in other words, by adaptively changing the number of search blocks for motion compensation according to the number of transmission frames per unit time, the hardware The capabilities of the software can be maximized, and the encoding efficiency is not significantly reduced, thereby solving the conventional problems.

The present invention was devised based on the inventor's knowledge, and is a processing amount adaptive type movement that can adaptively switch the processing amount according to the number of transmission frames per unit time. The purpose is to provide a compensated video encoding method.

[Means to solve the problem] FIG. 1 is a block diagram of the principle of the present invention.

Also in FIG. 1, in the transmitter 1, the input signal (
The motion compensation means 2 performs motion compensation on the original image information) by calculating a prediction error from, for example, the least squares error or absolute value error, and the jt quantization means performs quantization. .

By the way, the motion compensation means 2 is provided with search block number changing means 2A. Here, the search block number changing means 2A changes the number of search blocks for motion compensation according to the number of transmission frames per unit time.

[Operation] In the processing amount adaptive motion compensated video encoding method of the present invention described above, the screen is divided into blocks of a certain size for image information, and for each block, the prediction error is the smallest among the predicted screens. High-efficiency encoding is performed by searching for a block where . What is done is done. Thereby, the processing amount is adaptively switched according to the number of transmission frames per unit time.

[Example code] Hereinafter, the present invention will be described in detail with reference to the drawings.

FIG. 2 is a block diagram showing an embodiment of the present invention. In this FIG.
The motion compensation means 2 performs motion compensation by calculating a prediction error from the least squares error, absolute value error, etc., the quantization means 3 performs quantization, and the encoding means 4 encodes the prediction error. The input signal is sent to the receiving section 5.

Here, motion compensation refers to block i on the current screen [see Fig. 5(a)], block j on the prediction screen (for example, the previous screen in the case of interframe prediction)+Jz+'-・+
jn [see FIG. 5(b)] to detect a block with the smallest error determined by a certain evaluation method. Here, as an example of the evaluation method, there is an evaluation method using the sum of absolute differences. In this evaluation method, data (pixel values) in block i are divided into dk (k=1 to m,
m is the number of pixels in the block), and block ji (i=
l - n ) data in d k ' (k =
1-m), the evaluation value Si of block jj is obtained by.

Then, the block ji with the minimum Si becomes the optimal prediction block.

Further, to represent the block J1, a vector Vi (see FIG. 5(b)) indicating the positional deviation from the block 1 is used, and a plurality of detection target vectors are stored as a list.

Then, the vector corresponding to the above optimal prediction block is used as the optimal vector.

By the way, as shown in FIG. 3, the motion compensation means 2 includes a frame memory 21. Variable delay section 22. Motion compensation detection unit 2
It has 3.

Here, the frame memory 21 stores one frame of the previous screen as a prediction screen, and the variable delay unit 22 converts the image data to be fed back from the frame memory 21 into an optimal vector from the motion compensation detection unit 23. This is to delay by a predetermined amount based on the information.

The motion compensation detection section 23 finds the above-mentioned optimal vector from the original screen information and predicted screen information (previous screen information), and sends this optimal vector information to the receiving section 5 and also to the variable delay section 22. However, when determining the above-mentioned optimal vector, it has a function of changing the number of search blocks for motion compensation according to the number of transmission frames per unit time.

For this purpose, the motion compensation detection unit 23 uses the vector list 2
31. Evaluation value calculation circuit 232. Counter 233, comparator 234. A search block number setting section 235 is provided.

Here, the vector list 231 stores a plurality of detection target vectors v1 to Vn as a list, and uses a ROM, for example.

The evaluation value calculation circuit 232 calculates block j from data (pixel value) dk in block i of the original screen and data dk' in block jj of the predicted screen using equation (1) described above.
Calculate the block ji with the minimum evaluation value Si of i, and add the optimal vector corresponding to this block to the vector list 2.
31 and output to the variable delay section 22 and the receiving section 5.

The counter 233 functions as a pointer to a specific vector in the vector list 231, and
When 33 is cleared by a clear signal, it is incremented every time the evaluation value calculation circuit 232 performs an evaluation calculation on a certain block ji. In addition,
The count-up value from the counter 233 is input to the vector list 231 as well as to the comparator 234.

The comparator 234 compares the count signal from the counter 233 with the reference signal from the search block number setting unit 235 and calculates 1.
When the two become equal, a termination signal is issued.

The number of search blocks setting unit 235 changes the number of search blocks based on information on the number of transmission frames per unit time obtained from information such as the amount of storage in the buffer memory of the encoding means 4. When the number of transmission frames decreases, the number of search blocks is increased, and when the number of transmission frames increases, the number of search blocks is decreased.

With this configuration, the motion compensation detection unit 23 clears the counter 233 at the start of processing each block to output the first vector, and then increments the counter (+
1) t, te, vector list 231 stored in ROM
By sequentially specifying the vectors inside, the evaluation value S
The vector with the minimum i is searched for, but at this time, the maximum value (target block number; this is the number of transmission blocks) determined by the search block number setting unit 235 from the transmission speed or the number of bits per data The comparator 234 detects whether the value has been reached (a value that varies depending on the situation), and if it has been reached, a termination signal is issued and the vector corresponding to the one with the minimum evaluation value S1 at that point is selected as the optimal vector. Output as .

In this case, since the number of search blocks is changed according to the number of transmission frames per unit time, the amount of processing can be adaptively switched. In other words, when there is a large movement and the number of transmitted frames decreases, the number of search blocks is increased,
When the motion is small and the number of transmission frames increases, the number of search blocks is reduced, which allows the hardware to make full use of its capabilities without significantly reducing coding efficiency.

Of course, the functions of the motion compensation detection section 23 described above can also be realized by software processing. The processing procedure in this case is shown in FIG. In other words, the threshold value of the transmission rate or the number of transmission screens (the number of transmission frames per unit time) is set to ρ (rl~rΩ; r
l〉...〉rQ), first, in step AI, it is determined whether the number of transmission screens is larger than rl, and if it is larger, in step A2, the number of search blocks is set to 1, for example, and if it is smaller, For example, in step A3, it is determined whether the number of transmission screens is larger than r2, and if it is larger,
In step A4, the number of search blocks is set to 5, for example. After that, similar determinations and search block number settings are made, and finally, in step A5, it is determined whether the number of transmission screens is greater than rQ, and if it is, in step A6, the search block number is set. is set to 10, for example. If the number of transmission screens is not greater than rll, the number of search blocks is set to a value n greater than 10 in step A7.

After that, in step A8, i=1. After initialization with S=maximum value, in step A9, the evaluation value Si of vector v1 is calculated from the above formula (1), and if it is smaller than the previous minimum value, the minimum value is updated and a corresponding action is taken at the same time. The vector (or index) to be stored is stored (steps AIO to A13).

When the number of calculations determined at the start is completed, the yes route is taken in step A12, and the optimal predicted vector is determined in step A14.

In the receiving section 5, the received signal is decoded by the decoding means 6, motion compensated by the motion compensating means 7 in the same manner as the transmitting section 1, and further dequantized by the dequantizing means 8 to produce a reproduced image. It is designed to be output.

With the above configuration, high-efficiency encoding is performed by dividing the screen into blocks i of a certain size for image information, and searching for block j with the minimum prediction error from the prediction screen for each block i. However, the motion compensation means 2
The search block number changing means changes the number of search blocks for motion compensation in accordance with the number of transmission frames per unit time. This allows the amount of processing to be adaptively switched according to the number of transmitted frames per unit time.In other words, when the movement is large and the number of transmitted frames decreases, the number of search blocks is increased, and when the movement is small, the number of transmitted frames increases. Since the number of search blocks is sometimes reduced, the capabilities of the hardware can be maximized without significantly reducing coding efficiency.

Then, when the optimal vectors are obtained in this way and the prediction error signals of the input data are quantized, they are encoded and sent to the receiving section 5. The received signal containing the error signal information is decoded by the decoding means 6, the motion compensation means 7 performs motion compensation using the received optimal vector, and the dequantization means 8 further performs motion compensation using the received prediction error signal. After dequantization, the reproduced image is output.

[Effects of the Invention] As detailed above, according to the processing amount adaptive motion compensated video encoding method of the present invention, the number of search blocks for motion compensation is determined according to the number of transmission frames per unit time. Since the modification is performed, there is an advantage that the processing power of the hardware can be maximized without reducing the encoding efficiency.

[Brief explanation of drawings]

Fig. 1 is a block diagram of the principle of the present invention, Fig. 2 is a block diagram showing an embodiment of the present invention, Fig. 3 is a block diagram showing main parts of an embodiment of the present invention, and Fig. 4 is a block diagram of the present invention. 5(a) and 5(b) are diagrams illustrating the principle of motion compensation. FIG. 6 is a system configuration diagram illustrating the motion compensation method. In the figure, 1 is a transmitter, and 2 is a motion compensation means. 2A is a search block number changing means, 3 is a quantization means, 4 is an encoding means, 5 is a receiving section, 6 is a decoding means, 7 is a motion compensation means, 8 is a dequantization means, 21 is a frame memory, 22 23 is a variable delay unit, 23 is a motion compensation detection unit, 231 is a vector list, 232 is an evaluation value calculation circuit, 233 is a counter, 234 is a comparator, and 235 is a search block number setting unit.

Claims (1)

[Claims] Motion processing that performs highly efficient coding of image information by dividing the screen into blocks of a certain size and searching for a block with the minimum prediction error from the prediction screen for each block. A motion compensated video encoding method that is adaptive to processing amount and is characterized in that the number of search blocks for motion compensation is changed according to the number of transmission frames per unit time.