JPH0832046B2 - Variable rate image hierarchical coding method - Google Patents

Description

The present invention relates to a variable-rate image coding system using a high-speed packet network, which is a hierarchical code that reduces deterioration of image quality even if packets are discarded due to network congestion. It relates to the method of conversion.

"Prior Art" FIG. 4 is a diagram showing a transmission side of a variable rate image coding system for a communication network capable of varying a transmission rate such as a high-speed packet network. , Subtractor
In 101, the signal is subtracted from the signal of the previous frame stored in the frame memory 102, and only these difference signals are suppressed in redundancy by a coding method such as orthogonal transform coding and vector quantization in a coding unit 103. The information encoded by the encoding unit 103 is decoded by the decoding unit 104, added by the adder 105 with the signal of the previous frame read from the frame memory 102, and this signal is written in the frame memory 102. In this way, the frame memory 102 is updated with the latest information. The information encoded by the encoding unit 103 is also transmitted to the packet transmission unit 106, converted into a format suitable for a high-speed packet network in units of several blocks, and transmitted.

In this case, if the high-speed packet network ideally transmits coded information with a short delay regardless of the amount of generated information, the image quality becomes constant. However, when packet discard due to network congestion occurs, the difference signal between adjacent frames is encoded, and therefore the image after the discard is not reproduced accurately. To avoid this, it is conceivable to retransmit the discarded packet, but there are drawbacks that the delay time becomes long, the real-time property is lacking, and the network congestion further increases.

In order to solve this drawback, the image frame is divided into blocks, and the part that has a large effect on the image quality (hereinafter MSP: Most Significant Parts) and the part that has a small effect (hereinafter LSP: Least Significant Parts) Layered into several layers, MSPs and LSPs of several blocks are collected to form a packet consisting of MSPs and a packet consisting of LSPs only. At the time of transmission, a non-discardable identifier is discarded for packets consisting of MSPs and a packet is discarded for packets consisting of LSPs. A variable rate image coding method (Japanese Patent Application No. 63-165215) has been proposed in which a valid identifier is added, and when the network is congested, a packet consisting of an LSP to which a discardable identifier is added is discarded. Explaining this method, in the orthogonal transformation layered coding example, in the layering of MSP and LSP, the orthogonal transformation coefficient is scanned in a zigzag manner, the encoded information amount is obtained, and a preset value is uniformly set (for example,
Coefficients up to 75% of the total code amount per block are in MSP,
Other coefficients are assigned to the LSP, but if the packet containing the LSP of a block with a lot of movement is discarded due to network congestion, there is a situation in which it is not possible to restore an image without deterioration with the MSP code amount during decoding. On the other hand, there is a problem in that the MSP of a block having a small motion may be assigned with an excessive amount of code, and thus may have considerable redundancy.

An object of the present invention is to provide a variable rate image hierarchical encoding method that suppresses the influence of packet discard due to network congestion on image quality, which is a problem when transmitting a moving image signal in a high speed packet network.

"Means for Solving the Problem" The present invention has a great effect on the reproduction image quality of information obtained by orthogonally transform-coding a difference signal between adjacent frames (MS
P) and the one that has a small effect (LSP), and collects several blocks of MSPs and LSPs at the time of transmission to form a packet consisting of only MSPs and a packet consisting of only LSPs.
The packet consisting of is added with a non-discardable identifier indicating that it should not be discarded even during network congestion, and the packet consisting of LSP is added with a discardable identifier indicating that it may be discarded during network congestion. At the time, in the orthogonal transform layered coding transmission that discards only the packet consisting of the LSP to which the discardable identifier is added, in the layering, the power in the block is measured for each orthogonal transform block, and the power in the block is predetermined. If the block has an intra-block power equal to or higher than the threshold, the ratio of the code amount assigned to MSP to the total code amount in the block is increased to reduce the intra-block power below the threshold. For blocks that have, it is a technology whose main feature is adaptive orthogonal transform hierarchical coding that reduces the ratio of the code amount allocated to MSP to the total code amount in the block. The block code amount ratio to be assigned to the MSP to have set uniformly, by network congestion
When discarding a packet composed of an LSP, there is a possibility that the transmission code amount of a block having a large intra-block power may be insufficient, or the transmission code amount of a block having a small intra-block power may be excessive, and Is very different.

[Embodiment] FIGS. 1 (a) and 1 (b) are block diagrams of a transmission unit and a reception unit of an orthogonal transform hierarchical coding transmission method to which the present invention is applied, respectively. , Orthogonal transform encoder 203, layering unit 204, orthogonal transform decoder 205,
Adder 206, packet sending unit 207, packet receiving unit 208,
Orthogonal transform decoders 209, 210, frame memory 211, adder 21
It consists of 2,213.

The input image signal is subtracted from the signal of the previous frame stored in the frame memory 202 in the subtractor 201, and the difference signal is sent to the orthogonal transform encoder 203. The image signal encoded by the orthogonal transform encoder 203 is processed by the layering unit 204.
, The layer that has a large effect on image quality (MSP) and the part that has a small effect on image quality (LSP) are layered.
Only the SP is sent to the orthogonal transform decoder 205, and the image signal decoded in the orthogonal transform decoder 205 and the frame memory 2
The output of 02 is added in the adder 206, and the signal is stored in the frame memory 202 as the latest information. On the other hand, the MSP and LSP layered in the layering unit 204 are transmitted to the packet transmitting unit 207, and in the packet transmitting unit 207, the discarding identifier is assigned to the packet composed of the MSP and the discardable identifier is assigned to the packet composed of the LSP. Is added and sent to the high-speed packet network.

On the receiving side, the packet received from the high-speed packet network is discriminated between MSP and LSP in the packet receiving unit 208, and the MS
The P and LSP are sent to the orthogonal transform decoders 209 and 210, respectively. The signal decoded by the orthogonal transform decoder 209 is added to the signal stored in the frame memory 211 in the adder 212, and the signal is stored in the frame memory 211 as the latest information. On the other hand, the signal is sent to the adder 213, and the LSP arriving without being discarded in the orthogonal transform decoder 210 is added to the decoded signal to restore the moving image signal.

The present invention will be described in detail below with reference to FIGS. 2 and 3 by taking as an example the case where the intra-block power threshold for hierarchization is two.

FIG. 3 (a) shows a scanning method of the coding coefficient of one block after orthogonal transform coding. In the figure, one block is 8 × 8 pixels. It is also applicable to other block division methods such as 16 pixels. It is known that, when the layering unit 204 in FIG. 1 is layered into MSP and LSP, it is good to assign an 8 × 8 orthogonal transform coefficient to the MSP from the upper left part. Three
An example of scanning in the order shown by the arrow in FIG. 3A is shown, but the scanning is not necessarily performed in the order of FIG.

FIG. 2 is a block diagram showing the hierarchizing unit 204 in FIG. 1 in detail.
1, MSP code amount ratio input device 402, block power measuring circuit 403, block total code amount measuring circuit 404, hierarchical circuit 40
5 and up. Prior to transmission, an intra-block power threshold input device 401 and an MSP code amount ratio input device 402 are used to perform intra-block power thresholds pth1, pth2 and
Enter the MSP code amount ratio n1 (%), n2 (%), n3 (%),
Set.

In hierarchization, first, the power measurement device in the block 40
3. The in-block total code amount measuring circuit 404 calculates the in-block power p and the in-block total code amount c of each block. Next, the in-block power p is compared with two preset in-block power thresholds pth1 and pth2 (pth1 <pth2). If the in-block power p is (p ≦ pth1), The number of orthogonal transform coefficients to be assigned to the MSP is determined so that n1% of the total code amount c becomes the intra-block code amount of the MPS, and the remaining coefficients are assigned to the LSP. (FIG. 3 (b)).

Similarly, the power p in the block is (pth1 <p ≦ pth
2) and (pth2 <p), as shown in FIGS. 3 (c) and 3 (b),
Determine the number of orthogonal transform coefficients to allocate to MSP so that n2% and n3% are the intra-block code amount of MSP, and the remaining coefficients are LS
Assign to P. (However, n1 <n2 <n3) Further, for example, the distribution of the power p in the block is measured in advance, and the thresholds pth1 and pth2 are (p ≦ pth1) and (pth2 <p). If they are determined to be approximately equal, and n1, n2, and n3 are defined as n2 = (n1 + n3) / 2, the MSP transmission code amount is the transmission code when the MSP code amount is uniformly transmitted at n2%. Approximately the same amount, but the image quality in the event of LSP discard
As compared with the case where the code amount of 1 is uniformly transmitted at n2%, the image quality of a block having a large in-block power p is improved by the amount that the image quality is improved. (For a block with a small intra-block power p, only redundancy disappears and there is no deterioration in image quality.) In this embodiment, the intra-block power threshold value is set to two, but one threshold value is set. It is obvious that the above-described layering method and the layering method in which three or more threshold values are provided can be realized by the same technique.

“Effect of the Invention” As described above, when the moving image signal subjected to the orthogonal transformation is hierarchized, a threshold value is set for the intra-block power, and the intra-block power p is compared with the intra-block power threshold.
By adaptively changing (the amount of intra-block code assigned to MSP / the total amount of intra-block code), when the LSP packet is discarded, the transmission code amount is increased in a block with a large intra-block power to improve the image quality. On the other hand, in a block with low intra-block power, redundancy can be eliminated, so by setting the intra-block power threshold and the intra-block code amount ratio to appropriate values, the blocks allocated to conventional MSPs Compared to the case where the ratio of the inner code amount is uniform, there is an advantage that the image quality can be improved with the same code amount.

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of an orthogonal transform layered coding method according to the present invention, and FIG. 2 is a layering unit 20 in FIG.
4 is a block diagram showing a specific example, FIG. 3 is a diagram showing a method of layering orthogonal transform coefficients in one block for explaining the present invention, and FIG. 4 is a block diagram showing a conventional variable rate coding system. is there. 301: Orthogonal transform coefficient area including n1% of total code amount in block 302: Orthogonal transform coefficient area including n2% of total code amount in block 303: Orthogonal transform coefficient area including n3% of total code amount in block

 ─────────────────────────────────────────────────── ─── Continued Front Page (56) References JP 63-73786 (JP, A) IEEE TRANSACTIONS ON COMMUNICATIONS COM25 [11] P. 1285-1292

Claims (1)

[Claims] 1. An inter-frame differential signal of a moving image signal is orthogonally transformed and coded in block units, and the coded information is hierarchized into one having a large effect on reproduction image quality and one having a small effect on the reproduced image quality, and each layered. For each packet, several blocks are combined into one packet, and the packet that consists of only those that have a large effect on the playback image quality has an identifier that indicates that it should not be discarded even during network congestion. In a variable rate image hierarchical coding method in which a packet consisting of only one is added with an identifier indicating that it may be discarded at the time of network congestion and is packet-multiplexed and transmitted, one with a large impact on playback image quality and one with a small impact on playback image quality When hierarchizing, a threshold value is set in advance in the power in the block subjected to the orthogonal transformation, and the power in the block is measured for each block. Due to the magnitude relationship between the intra-block power and the threshold value, in the block with high intra-block power, the ratio of the code amount allocated to the one that has a large effect on the playback image quality to the total intra-block code amount is increased, Then, a variable rate image hierarchical coding method characterized by a hierarchical structure in which the ratio of the code amount assigned to the one having a great influence on the reproduction image quality to the total code amount in the block is reduced.