JPH0215791A - Variable-rate picture encoding system - Google Patents

Abstract

PURPOSE:To reduce deterioration in picture quality caused by abandonment of packets by hierarchizing information obtained by encoding moving image signals into parts largely affecting the picture quality and parts little affecting the picture quality so that each part can be identified and calculating difference signals between adjacent frames from decoded signals of the largely affecting parts only. CONSTITUTION:Difference signals between inputted moving image signals and signals of the previous frame accumulated in a frame memory 2 are encoded and hierarchized into parts (MSP) largely affecting the picture quality and parts (LSP) little affecting the picture quality. The MSP is decoded and used for rewriting the memory 2 after the MSP is mixed with signals read out from the memory 2. At the same time, the MSP and LSP are respectively gathered by frames or blocks and sent in a packet multiplexed state after abandonment disable identifiers and abandonment enable identifiers are respective added to the MSP and LSP. On a receiving side, the encoded signals are separated into the MSP and LSP based of the abandonment enable/disable identifiers after receiving the packets and signals obtained by decoding the MSP and the signals of the previous frame accumulated in a frame memory 11 are added to each other. After adding the signals, signals obtained by decoding the LSP are added to the added signals and moving images are obtained.

Description

DETAILED DESCRIPTION OF THE INVENTION "Field of Industrial Application" This invention is a variable rate image encoding method (== that uses a high-speed packet network) that eliminates the influence of image quality deterioration even if packets are discarded due to halftone width. This relates to an encoding method that reduces

"Prior Art" In general, high-efficiency image encoding systems encode difference signals between adjacent frames of consecutive image frames. Therefore, if the movement is large, the amount of information generated is large, and if the movement is small, the amount of information generated is small.

In order to transmit this encoded information via a communication channel with a constant transmission speed (- means that if a buffer memory is set up and the transmission speed is constant, the encoded information stored in the buffer memory will be a constant amount. There was a drawback that the encoding parameters had to be controlled and the image quality varied.

FIG. 3 is a diagram for explaining a variable rate image encoding method for communication networks that can vary the transmission speed, such as high-speed packet networks.
1 from the previous frame signal stored in the frame memory 32, and only the difference signal is sent to the encoding section 33. In the encoding unit 33, redundancy is suppressed using well-known encoding methods such as orthogonal transform encoding and vector quantization.

The information encoded by the encoding unit 33 is decoded by the decoding unit 34, added to the signal of the previous frame read from the frame memory 32 by the adder 35, and this signal is written to the frame memory 32. In this way, the frame memory 32 is updated to the latest signal. The information encoded by the encoding unit 33 is also sent to the packet sending unit 36, where it is converted into a format compatible with high-speed packet networks every image frame, or every two blocks when one surrounding frame is divided into blocks. and sent out.

In this case, ideally, if a high-speed packet network can transmit encoded information with a short delay regardless of the amount of information generated, the image quality will be constant.However, if packets are discarded due to the network width @42, , since the difference signal between adjacent frames is encoded, images after discarding will not be accurately reproduced.

To avoid this, it is possible to retransmit the discarded packets, but this increases the delay time and lacks real-time performance.
Furthermore, there was a drawback that the net width became even larger.

SUMMARY OF THE INVENTION An object of the present invention is to provide a variable rate image encoding system that suppresses the influence of packet discard due to network width @(-) on image quality, which is a problem when transmitting moving image signals over a high-speed packet network.

"Means for Solving the Problems" This invention divides information obtained by encoding the difference signal between adjacent frames into a part that has a large influence on image quality (hereinafter abbreviated as MSP) and a part that has a small influence (hereinafter abbreviated as LSP). ), and the MSP cannot be discarded, and the LSP has an identifier that can identify the discard town, and when the network is wide, only LSPs with a discardable identifier are discarded. The characteristics are as follows. Therefore, MSP and LSP are sent to the high-speed packet network without distinction, and when the network width is full, MSP and LSP are discarded without distinction, which is very different from the technology of the Congregational Song Dynasty, where image quality deteriorates significantly.

Embodiment FIG. 1 is a detailed diagram of the present invention, in which 1 is a subtracter, 2 is a frame memory, 3 is an encoding section, 4 is a layering section, 5 is a decoding section, and 6 is a diagram for explaining the present invention in detail. adder, 7 is a packet sending unit,
8 is a packet receiving section, 9.degree. 10 is a decoding section, 11 is a frame memory, and 12.degree. 13 is an adder.

To operate this, first, the input moving image signal is subtracted from the signal of the previous frame stored in the frame memory 2 (-) by the subtracter l, and only the difference signal is sent to the encoding unit 3. Encoding The information encoded by the section 3 is divided into two layers by the layering section 4: a portion having a large influence on image quality (MSP) and a portion having a small influence on image quality (LSP).

The layering method will be explained in detail with reference to FIG. 2, focusing on the orthogonal transform encoding method. The difference signal between adjacent frames from the subtracter 1 is orthogonally transformed for each block of 8 pixels x 8 lines, as shown in FIG. 2(a), for example.

The orthogonal transformation coefficients can be expressed in a matrix as shown in Figure 2 (b) (-), and the coefficients in the upper left direction include the low frequency components of the signal and are the part (MSP) that has a large effect on the image quality (-), and the coefficients in the lower right direction The direction coefficient determines the image quality (including high frequency components of the signal).
=Low influence part (LSP). To layer MSP and LSP, as shown in FIG.
The coefficients up to 5%) are used as MSP, and the other coefficients are used as LSP. Another method for layering is to add the power of each coefficient scanned in a zigzag manner, and set the coefficients up to a preset value as MSP and the other coefficients as LSP. In this case, the SN ratio of each block is constant. Note that the coefficient scanning method is an example.
Other methods may also be used.

Among the hierarchical encoded information, MSP is decoded by the decoding unit 4, and added to the signal decoded only from the MSP of the previous frame read from the frame memory 2 by the adder 6, and this signal is added to the frame Write to memory 2. Like this C
The frame memory 2 is then updated with the latest signal C2.

The MSPs and LSPs that have been hierarchized by the hierarchization unit 4 are grouped by the packet sending unit 7 for each MSP and each LSP, and are sent out with headers necessary for high-speed packet network transmission added. M.S.
A non-discardable identifier is added to the header C2 of the packet consisting of P, and a discardable identifier is added to the header of the packet consisting of LSP. The unit in which MSP and LSP are grouped may be a frame unit, a block unit for orthogonal transform encoding, or an integral multiple of a block. In this case, the packet length is variable, but ATM, which is currently being considered for standardization by CCITT as a high-speed packet network,
Asynchronous Transfer Mod
e) When a network is applied, it becomes a fixed length/short packet (called a cell). When transmitting on an ATM network, each MSP and L
After arranging them for each SP, they may be divided according to cell length, and a non-discardable identifier may be added to the header of a cell consisting of MSPs, and a discardable identifier may be added to the header of a cell consisting of LSPs.

On the receiving side, the packet sent from the transmitting side via the high-speed packet network is first decoded by the packet receiving section 8, and the header is decoded, and the packet is separated into MSP and LSP4. After the MSP is decoded by the decoder 9, the adder 12 adds it to the signal read from the frame memory 11, sends it to the adder 13 (-rewrites the frame memory 11, and separates it by the packet receiver 8). The decoded LSP is decoded by the decoding unit 10, and then sent to the adder 13, and the MSP is added to the decoded signal to obtain an output moving image signal.

With such a configuration, when the high-speed packet network becomes unstable, only the LSP can be discarded using the discardable/discardable identifier C2.

As a result, even if packets are discarded, image quality deterioration is significantly reduced compared to conventional techniques.

In the above explanation, orthogonal transform was used as the encoding method, but the present invention can be applied to other encoding methods, such as vector quantization or a method that combines motion compensation, as long as it can be layered into MSP and LSP. .

"Effects of the Invention" As explained above (-According to the present invention, information obtained by encoding a moving image signal is hierarchically divided into parts that have a large effect on image quality and parts that have a small effect on image quality, and each part can be identified. Since the difference signal between adjacent frames is calculated from the decoded signal of only the portion that has a large influence, there is an advantage that image quality deterioration due to packet discard can be reduced and shortened.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is a diagram explaining an orthogonal transform encoding method, and FIG. 3 is a block diagram explaining a conventional encoding method. Patent applicant: Nippon Telegraph and Telephone Corporation Representative: Kusano In the car 1
Circle ke

Claims (1)

[Claims] (1) means for encoding a difference signal between an input moving image signal and a signal stored in a frame memory; and means for hierarchizing the encoded information into parts that have a large effect on image quality and parts that have a small effect on image quality; A means for decoding the information of the part with a large influence and rewriting the frame memory with a signal mixed with the signal read from the frame memory, and a means for rewriting the information of the part with a large influence and a part with a small influence in the hierarchy, respectively. A means for collecting the packets in frames or blocks, adding a non-discardable identifier to the information of a portion having a large influence, adding a discardable identifier to the information of a portion having a small influence, multiplexing the packets, and transmitting the packets; means to receive the information and separate it into the information of a large impact part and the information of a small part by a discardable/disposable identifier, and a signal decoded from the information of the large impact part and a signal of the previous frame stored in a frame memory. A variable rate image encoding method comprising means for obtaining a moving image signal by adding the information of the portion having a small influence with the decoded signal after adding the information of the signal and the decoded signal.