JPH0813137B2 - Variable rate video coding device - Google Patents

Description

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a variable rate moving image coding device used in a videophone, a video conference system, and the like.

2. Description of the Related Art FIG. 4 shows a schematic configuration of an example of a conventional variable rate moving image coding apparatus. In FIG. 4, a moving image signal
40 is input in frame units. On the other hand, the prediction signal of the current frame is obtained by performing motion compensation or the like on the previous frame reproduction signal stored in the frame memory 48. The obtained prediction signal and the moving image signal 40 are subtracted to form a prediction error signal, and hereinafter, processing is performed for each block of size N × N.

The N × N prediction error signal is discrete cosine transformed in the discrete cosine transform circuit 41 and then input to the quantization circuit 42. The quantization circuit 42 quantizes the coefficient that has been subjected to the discrete cosine transform, and the quantized representative value is input to the layering circuit 43. The hierarchization circuit 43 includes a portion (hereinafter, abbreviated as MSP) including a low frequency component of the prediction value error signal and a portion (hereinafter, referred to as LSP) including the low frequency component of the N × N quantized representative value. The MSP is transmitted on the variable speed high priority channel 44, and the LSP is transmitted on the variable speed low priority channel 45. The variable speed high priority channel 44 and the variable speed low priority channel 45 are both variable speed channels capable of supporting arbitrary transmission speeds, and are realized by the broadband ISDN currently being studied as a next-generation communication network. Wideband ISDN provides multiple variable speed channels with different priorities,
When the communication network (hereinafter, abbreviated as network) is congested, it is planned that the control will be performed so that the packet of the channel with a low priority is discarded. Therefore, variable speed high priority channels
44 is a channel with a high priority, that is, a packet in which packets are not discarded even when the network is congested, and a variable speed low priority channel 45 is a channel in which packets may be discarded when the network is congested.

Local decoding is performed using only MSP. For decoding on the receiving side, MSP adds the decoded signal of LSP after adding it to the value of the previous frame.

Therefore, the LSP is stochastically discarded when the network is congested, but the LSP itself has a small influence on the image quality, and the deterioration of the image quality can be suppressed because it does not spread over time.

However, in the conventional example, the ratio of the code amount of MSP and LSP, or the S / N ratio in the block is used as a threshold value for MSP and LSP.
However, there is a problem that the effect of image quality deterioration when LSP is discarded on visual perception is not taken into consideration.

The present invention solves the above problems, and provides a variable rate moving image coding apparatus capable of guaranteeing a certain preset image quality in consideration of visual characteristics even when an LSP is discarded. The purpose is to do.

Means for Solving the ProblemsThe present invention, in order to achieve the above-mentioned object, weights a quantization error by a visual characteristic, and uses a weighted quantization error to separate MSP and LSP. .

Effect The present invention, by the above configuration, when the network is congested, discards only the LSP that is visually unnoticeable even if it is discarded, and also considers the visual characteristics when the LSP is discarded. It is possible to guarantee a certain level of image quality with MSP.

Embodiments Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram showing an embodiment of a variable rate moving image coding apparatus according to the present invention, FIG. 2 is an explanatory diagram showing a scanning order when determining a separation boundary between LSP and MSP, and FIG. It is explanatory drawing which shows the weighting coefficient for weighting the quantization error in a discrete cosine transform domain.

In FIG. 1, 10 is a moving image signal, 11 is a discrete cosine transform circuit for performing discrete cosine transform of an inter-frame prediction error signal in block units of size N × N, and 12 is a coefficient transformed by the discrete cosine transform circuit 11. Is a quantization circuit that quantizes. 13 is LSP / MSP that separates each block into MSP and LSP
It is a separation circuit. Both 14 and 15 are variable-rate channels that can support arbitrary transmission rates, 14 is a high-priority channel, that is, a variable-rate high-priority channel in which packets are not discarded even when the network is congested, and 15 is a network. This is a variable rate low priority channel where packets may be dropped during congestion. Reference numerals 16 and 17 denote an inverse quantization circuit and an inverse discrete cosine transform circuit for local decoding. Reference numeral 18 is a frame memory for storing reproduced images, and 19 is a prediction coefficient multiplier.

Next, the operation of the above embodiment will be described. In FIG. 1, the moving image signal 10 is input in frame units. On the other hand, the prediction signal of the current frame is obtained by subjecting the reproduction signal of the previous frame stored in the frame memory 18 to motion compensation and the like, and then multiplying by the prediction coefficient α (0 <α ≦ 1) in the prediction coefficient multiplier 19. To be The obtained prediction signal and the moving image signal
10 is subtracted to form a prediction error signal, and thereafter, processing is performed for each block of size N × N.

The N × N prediction error signal is discrete cosine transformed in the discrete cosine transform circuit 11, and then input to the quantization circuit 12. The quantizing circuit 12 quantizes the coefficient subjected to the discrete cosine transform, and the quantized representative value is input to the LSP / MSP separating circuit 13 and the inverse quantizing circuit 16 for local decoding.
Is also entered. The coefficient inversely quantized by the inverse quantization circuit 16 is inverse discrete cosine transformed by the inverse discrete cosine transform circuit 17, then added to the prediction signal, and stored in the frame memory 18.

The LSP / MSP separation circuit 13 uses the transform coefficient that is the output of the discrete cosine transform circuit 11 and the transform coefficient that has been quantized by the quantizer circuit 12 to make LSP and MSP
To separate.

Next, in the LSP / MSP separation circuit 13,
The method of separating into SP and MSP is explained.

The size of the block as the processing unit is N × N, and the transform coefficient output from the discrete cosine transform circuit 11 is D (u,
v), the transform coefficient after quantizing this D (u, v) by the quantizing circuit 12 is Dq (u, v), and its quantized representative value is Ix (u, v). Further, the weighting coefficient considering the visual characteristics in the discrete cosine transform domain is W (u, v). However, u = 0,1,2, ..., N-1 v = 0,1,2, ..., N-1.

First, when the LSP is discarded due to network congestion, the minimum guaranteed intra-block error (the sum of the quantization error generated by the quantization of the coefficients in the MSP area and the error generated by the LSP being discarded) The root mean square value WMSeth is set in advance.

Next, for each block, Moreover, the minimum (ζ, η) that satisfies WMSE ≦ WMSEth (2) is obtained, and this point is set as the separation point of LSP / MSP.

However, the scanning of u and v in the equation (1) is performed from the low frequency component to the high frequency component as shown in FIG. 2, for example.

As described above, according to the above-described embodiment, when the network is congested, the LSP
Even if the packet composed of is discarded, MSP can guarantee the root mean square of the intra-block error to be equal to or less than the preset WMS Eth. Also, since the LSP is discarded, there is a difference in image data between transmission and reception, but this effect decreases with time due to the prediction coefficient α.

In the above embodiment, the scanning method as shown in FIG. 2 is used as the scanning order of u and v when calculating the equation (1), but the present invention is not limited to such scanning order. . Although the numerical values shown in FIG. 3 are given as examples of the weighting coefficient of the visual characteristic in the discrete cosine transform area, the present invention is not limited to these numerical values.

As described above, according to the present invention, the intra-block error is weighted by the visual characteristic, and the MS of the weighted intra-block error is set to be smaller than a preset threshold value.
Since P and LSP are separated, even if the packet composed of LSP is discarded when the network is congested, MSP can guarantee a certain preset image quality in consideration of visual characteristics. .

[Brief description of drawings]

FIG. 1 is a block diagram showing an embodiment of a variable rate moving image coding apparatus according to the present invention, FIG. 2 is an explanatory diagram showing a scanning order when determining a separation boundary between LSP and MSP, and FIG. Explanatory drawing showing a weighting coefficient for weighting a quantization error when calculating a mean square error in a discrete cosine transform region,
FIG. 4 is a block diagram showing an example of a conventional variable rate moving image coding apparatus. 10 ... Moving image signal, 11 ... Discrete cosine transform circuit, 12 ...
… Quantization circuit, 13 …… MSP / LSP separation circuit, 14 …… Variable speed high priority channel, 15 …… Variable speed low priority channel, 16 …… Inverse quantization circuit, 17 …… Inverse discrete cosine transform circuit , 18 ... Frame memory, 19 ... Prediction coefficient multiplier.

Claims (1)

[Claims] 1. A predictive error signal is divided into blocks of size N × N, and in the inter-frame predictive coding of a moving image signal for which discrete cosine transform is performed for each block, the discrete cosine transform of N × N is performed. The quantization error of each transform coefficient when the number of transform coefficients is quantized is weighted by visual characteristics, and N × N quantized representative values are used for the low frequency of the prediction error signal using the weighted quantization error. Separation means for separating into a region including a component and a region including a high frequency component of the prediction error signal is provided, and the region including the low frequency component is transmitted by a variable speed channel having a high priority and includes the high frequency component. A variable rate moving picture coding apparatus adapted to transmit a region on a variable speed channel having a low priority.