JPH02248179A - Image encoding method - 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 [Object of the Invention] (Field of Industrial Application) The present invention relates to an image encoding system for video conferences, video telephones, and the like.

(Prior Art) FIG. 2 is an example of a conventional moving image encoding apparatus for the above-mentioned video conference or the like. A motion vector MY is obtained for the motion compensation calculation circuit that compares the image signal human block with the block of the previous frame that has been translated in parallel and finds the block of the previous frame with the smallest error. Filtered by filter 112 and subtractor 1
01, the difference with the human block is taken. Comparison judge 1
02 input (no-MC) i-difference (MC
) is made, and the selection information is sent to the multiplexer 106.
sent to. The information amount of the selected information is calculated by the comparator 103 and compared with a predetermined threshold. When the information is larger than the threshold, information indicating that it is a coded block (Co
ded) is sent via multiplexer 106. When the value is smaller than the threshold value, information indicating that the block is not a coded block (non-Coded) is sent to the multiplexer 106.
sent via. The encoded block is cosine transformer 1
The signal is cosine-transformed at step 04, quantized at quantizer 105, and sent to multiplexer 106. The data dequantized by the dequantizer 107 is inversely transformed by the inverse cosine transformer 108, added to the data used for motion compensation by the adder 109, and stored in the frame memory 110. For motion compensation not performed (no-MC block), the signal applied to the adder 109 by the changeover switch 113 is O, and this adder passes the DCT data as is to the frame memory 110. Sent. In these operations, the loop filter 112 has a motion vector of O
It is also possible to perform control in which the signal is turned off when the signal is a vector, and turned on at other times.

A method for adding encoding control information in such a device will be explained.

FIG. 3(a) is an example of dividing an image into large blocks, medium blocks, and small blocks, and FIG. 3(b) is an example of medium blocks. Large block is 352x288 pixels, small block size is 8
Cosine transformation is performed in this unit of ×8 pixels. In this encoding, the quantization step size is set once for each large block, and can also be set for each medium block according to instructions in the header information. FIG. 4 shows an example of a format for transmitting quantization step size information and quantized data after cosine transformation. There is a code Ql indicating the quantization step size in front of each block, and a code Q2 that allows the quantization step size and its identification is set in front of each medium block in the large block as necessary. ing.

As shown in Figure 3('b), four small blocks and two medium blocks are constructed, and one small block is 8×
It is composed of an 8-pixel image, and the motion-compensated data is cosine-transformed in block units, quantized with the quantization step size set at that time, and variable-length encoded.

(Problem to be Solved by the Invention) However, in the conventional encoding device described above, since the amount of generation of the code Q2 of the quantization step size for the medium block is not checked in advance, the conversion of the quantization step size may be necessary depending on the input image. There is a problem in that the amount of step size code that occupies a code word increases abnormally. Particularly in video transmission at low bit rates, such as 64 kbps or 128 kbps, it is important to reduce the overhead information occupied in code words as much as possible. Significant deterioration in image quality occurred, that is, data occupied by code words, resolution, gradation, and number of frames decreased. In addition, if the quantization step size is inserted only at a certain position, the image quality may deteriorate if it cannot correspond to the detailed position of the image, and the encoded diameter information may be lost on the transmission path. When the bit rate is small, the quantization step size cannot be precisely controlled, resulting in inflexibility.

Therefore, an object of the present invention is to provide an image encoding method that optimizes image quality by controlling the amount of information generation suitable for the transmission bit rate given the quantization step size.

[Structure of the invention]

(Means for Solving the Problem) In order to solve the above problem, the present invention monitors the total amount of overhead in a medium block, and when the amount increases, prohibits the occurrence of overhead, and thereafter overhead information is used. The monitoring method includes, for example, switching control that allows or prohibits the addition of overhead information based on the total number of generated bits of overhead information in a medium block, the number of times overhead information has been added, and whether or not it has occurred continuously. This is what we do.

(Function) According to the present invention, it is possible to suppress the overhead information that occupies the code amount after image encoding to a certain level or less, and it is possible to prevent the image quality from greatly decreasing due to excessive overhead. It is possible to perform balanced rate control by separating the contents of different codes from the simple rate control using a conventional output buffer.

(Example) FIG. 1 shows an example of the present invention. The configuration is similar to the conventional example shown in FIG. 2, with the addition of a determination circuit that controls changes in step size and outputs determination results. A block diagram of this determination circuit is shown in FIG. In the video multiplex circuit 106, motion compensation on10ff data is coded or non-coded.
Data indicating the motion vector, motion vector information, etc. are multiplexed and sent to the output buffer. Among them, an example of encoding a medium block and a quantization step size for the large block has already been explained with reference to FIGS. 3 and 4. Conventionally, the step size has been changed depending on the size of the output buffer amount. That is, when the buffer amount increases, the step size is increased, and when the buffer amount decreases, the step size is decreased. Here, in order to limit the number of times the step size is changed, the number of times the step size is changed in one large block is counted by a counter 151, and when the number exceeds a certain value, changing the step size is prohibited. Thereafter, the same step size as before the change is used. This counter 151 is reset to 0 at the beginning of a large block, and its output is compared with a preset value in a comparator 150, and when the value exceeds the value, unchangeable information such as O is output. Until then, changeable information such as l is output.

In this embodiment, the number of step size changes is checked using an up counter and a reset, but a down counter may be used as long as the number of times the step size is changed is checked.

Furthermore, after changing the step size is prohibited, even if the amount of information generated increases or decreases, the step size cannot be controlled until the start of the next large block, so the final step size when the step size is fixed is relatively small. It is safer to fix it to a large value because overflow will not occur. In the case where the determination result of the comparator 150 is changeable, an embodiment of which is shown in FIG.
The input step size is used as is and becomes the output. When the determination result is in an unchangeable state, the switch is connected to the lower side, and the input step size is changed to a value of 64 and output.

FIG. 7 shows another embodiment of the determination circuit. Every time the step size is updated, the code amount when encoding the new step size is read out from the code table memory 155, and the sum is taken by the arithmetic circuit 1.56 that calculates the total code amount. When the value exceeds the value, the value cannot be changed. In this example, when the number of types of code amounts of the quantization step size is small, the code amount summation circuit 156 may be a simple counter. Another embodiment is shown in FIG. 8. This prohibits updating when the step size is updated twice in a row. The value of the 2-bit counter 160 increases by 1 in the middle block where there is an update, and when there is no update, this counter is reset to °゛OO''. After one update, it becomes °01! When updated continuously”
01''! The 12th bit is set to 1. This is taken into the latch 161 and output as unchangeable information. This latch 161 is reset at the start of a large block.

Another embodiment is shown in FIG. In contrast to the conventional example shown in Figure 2,
The difference is that the generation of motion vectors is suppressed depending on the amount of motion vectors generated. Conventional motion vector is 16x16
Calculate once for a medium block of pixels, and then calculate four for each small block of 8 x 8 pixels, and if the difference from the motion vector of the medium block is large, use the necessary amount of small block motion vectors, that is, 1 to 4. It is sent with overhead information indicating that it was sent. FIG. 10 shows a code example in that case. All four small blocks in the first medium block use the same motion vector MV1, and the first two small blocks in the second medium block use the same motion vector MV1. While the motion vectors indicated by MVI and MV2 are used, only the motion vector MVI is used in the subsequent two small blocks.

In such a configuration, when the amount of motion vectors generated for small blocks begins to increase, especially in low bit rate moving image transmission, the amount of motion vector information increases and the image quality deteriorates significantly. Therefore, the determination circuit 202 monitors the occurrence of motion vectors, and when the increase exceeds a certain limit, the transmission of motion vectors in new small blocks is restricted, and the switch 201 performs fixed motion compensation. It is. It can be seen that if the motion vector MVI in units of medium blocks is excluded because it is always sent, only the motion vector MV2 in units of small blocks needs to be counted. The code amount of this MV2 is given to the MV code amount memory 210, summed by a counter 211 for calculating the sum, and compared with a fixed value in 212 to control motion vector generation in units of small blocks.

If the actual motion vector is a variable length code, this motion vector code amount is used to generate the number of variable length code bits. Furthermore, when the calculation is to be simplified or when the motion vector is a fixed length code, it is sufficient to simply generate the number of occurrences of MV2.

Further, the operation of the determination circuit 202 can be controlled depending on whether MV2 has occurred continuously.

In the above embodiment, the motion vector was calculated in two steps: a medium block of 16 x 16 pixels and a small block of 8 x 8 pixels. It can also be done in units of li. That is 8
4 motion vectors MV 2A, MV 2B in ×8 units
, MV 2 C, and MV2D are determined, and if the determination circuit 202 permits, all four motion vectors are transmitted; if prohibited, the representative motion vector MV2' is set to an average vector such as the formula 0, or a vector such as the formula (■). It is possible to suppress the amount of motion vectors generated by making one of them the representative. Further, in this suppression, the motion vectors of small blocks may be limited in stages from 4 to 3 to 2 to 1.

〔Effect of the invention〕

As described above, according to the present invention, it is possible to optimize the image quality by controlling the amount of information generation suitable for the transmission bit rate given the quantization step size.

[Brief explanation of drawings]

FIG. 1 is a diagram showing an embodiment of the present invention, FIG. 2 is a diagram showing a conventional example, FIGS. 3 to 4 are diagrams showing an example of encoding the quantization step size, and FIG. 5 is a determination circuit. Block diagram, No. 6
9 to 9 are diagrams showing other embodiments of the present invention, and FIGS. 10 to 12 are diagrams for explaining encoding. Agent Patent Attorney Noriyuki Ken Yudo Matsuyama Hikaru = 52 (string)

Claims (1)

[Claims] Encoding control information is added to input image data in blocks of different sizes and sequentially encoded, and the next encoding control information is added based on this encoded output. An image encoding method characterized by determining the size of blocks.