JPH07322267A - Image signal coder - Google Patents

Abstract

PURPOSE:To attain quantization control adaptively for a local characteristic of an image in the image high efficiency coding. CONSTITUTION:An image signal received from an input terminal equipment 101 is divided into small blocks by a small block division section 102. A characteristic quantity extract section 108 extracts plural characteristic quantities for each small block and a characteristic quantity distribution state calculation section 109 calculates a distribution state of the entire image for each characteristic quantity. A weight calculation section 110 uses the distribution state of each characteristic quantity to calculate a weight coefficient for each characteristic quantity. A quantization control parameter calculation section 111 uses the calculated weight coefficient and plural characteristic quantities calculated for each small block to calculate a quantization control parameter. A quantization step calculation section 112 decides a quantization step of a quantization section 104 based on the quantization control parameter to be calculated and an occupied state of a buffer 106.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image signal coding apparatus, and more specifically, in high efficiency coding of a television or the like, a limited amount of information is adaptively allocated according to a local property of an image. Accordingly, the present invention relates to an image signal encoding device for improving image quality.

[0002]

2. Description of the Related Art Conventionally, in high-efficiency coding of an image signal of a television or the like, an area where coding noise is easily detected is quantized by a fine quantization step, and an area where coding noise is hard to be detected is coarsely quantized. The method of quantizing in steps is often used.

FIG. 3 shows a configuration example of conventional quantization control by dispersion of image signals. In this example, it is assumed that standard motion and discrete cosine transform are used as the image coding method. Also, as one of the human visual characteristics, there is a masking effect that a flat area with little change will be noticed if deterioration occurs even a little, and a region with a lot of change has a masking effect even if it deteriorates a little. The variance of pixel values within a frame is used as a feature amount for switching.

First, the image signal input from the input terminal 301 is divided by the small block dividing unit 302 into N × N small blocks. A motion vector 305 is obtained by the motion vector detection unit 304 from the image signal 303 of each small block and the image signal of one frame before stored in the frame memory 314. Based on the motion vector 305, the motion compensation unit 306 performs motion. After the compensation is performed and the motion compensation prediction error signal 308 is obtained by the subtractor 307, the motion compensation prediction error signal 308 is subjected to the discrete cosine transform in the discrete cosine transform unit 309.

On the other hand, the image signal 30 divided into small blocks
For 3, the intra-frame variance (V) 316 is obtained by the variance calculation unit 315 by the following calculation and is sent to the quantization step calculation unit 317.

[0006]

[Equation 1]

In the quantization step calculation unit 317,
Based on the intra-frame variance (V) 316 and the buffer memory occupancy 318, under certain information, when the intra-frame variance 316 is large, a coarse quantization step is set, and the small part is finely set. Done. Using the set quantization step 319, the discrete cosine transform coefficient 320 is quantized by the quantizer 310, variable-length coded by the variable-length coding unit 321, and then input to the buffer memory 322, where it is kept constant. Is output to the encoded data output terminal 323 at the bit rate of.

The discrete cosine transform coefficient quantized by the quantizer 310 is inversely quantized by the inverse quantizer 311 and inverse discrete cosine transformed by the inverse discrete cosine transform unit 312, and then by the motion compensator 306. The data of the motion-compensated previous frame is added by the adder 313 and stored in the frame memory 314 for motion-compensated prediction of the next frame.

According to the configuration as shown in FIG. 3, since the flat portion can be finely quantized and the portion having a large change can be roughly quantized, a limited amount of information can be distributed according to the masking effect. As a result, the quality of the encoded image can be improved.

[0010]

SUMMARY OF THE INVENTION In the above conventional method,
Since the quantization control is performed using only one type of feature quantity,
It was a problem to decide what kind of feature quantity to use.

For example, as shown in FIG. 4, let us consider a case where the intra-frame variances are substantially equal, and a region (a) in which the motion is uniform and a random region (b) coexist. When intra-frame separation is used as a feature amount of quantization control for such an image signal, the fineness of quantization is controlled to be substantially equal in both (a) and (b). In this case, the deterioration is more noticeable in the area where the movement is uniform as compared with the area where the movement is random. In such a case, it is considered appropriate to make the quantization step of (a) finer than that of (b), but (a) and (b) cannot be distinguished only by intraframe dispersion. , (A) the quantization step is not controlled in detail.

On the contrary, consider the case where the magnitude of the difference vector between the motion vectors of the adjacent small blocks is selected as the feature quantity representing the randomness of the motion. As shown in FIG.
If the entire screen is moving in parallel, the quantization steps in all areas will be controlled equally.

As described above, the conventional quantization control using only one type of feature quantity has a problem that the control does not work properly depending on the image signal.

An object of the present invention is to solve the above-mentioned problems of the prior art, and to improve the image quality as compared with the prior art by adaptively performing the quantization control for the local feature of the image. An object of the present invention is to provide an image signal encoding device that enables the above.

[0015]

An image signal coding apparatus according to the present invention comprises means for dividing an image signal of a current frame into small blocks, means for obtaining a plurality of feature quantities for each small block,
Means for obtaining a distribution state for each feature amount over one frame, means for weighting each feature amount according to the obtained distribution state for each feature amount, and a plurality of features obtained for each small block Means for determining one new feature amount by weighting the calculated amount with the calculated weight, and quantizing the encoding target small block using the one calculated feature amount. It is characterized by switching characteristics.

[0016]

According to the present invention, the weighting coefficient of the feature amount for each small block is obtained from the distribution state of the feature amount of the entire screen, and only the effective feature amount among the plurality of feature amounts is automatically subjected to the quantization control. To reflect.

[0017]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described below with reference to the drawings.

FIG. 1 shows the basic configuration of the image signal coding apparatus of the present invention. Here, the number of feature quantities used is n.
The image signal input from the input terminal 101 is divided into small blocks by the small block dividing unit 102, and the image signal 11 is divided by the feature amount extracting unit 108 for each divided small block.
3 to a plurality of feature quantities f _i (i = 1, 2, ..., N) 11
Extract 4. Next, in the feature quantity distribution state calculation unit 109, for each feature quantity f _i , the distribution state m _i (i =
1, 2, ..., N) 115 is calculated. Using the distribution state m _i of the feature amount, the weight calculation unit 110 calculates the weight coefficient w _i (i = 1, 2, ... N) 116 for each feature amount from the following equation.

[0019]

[Equation 2]

Here, W _i is calculated from the distribution state of each feature
It is a function that increases the weighting coefficient w _i of each feature quantity that is effective for the quantization control of the image to be encoded and decreases the weighting coefficient w _i of the other feature quantities.

Subsequently, the weight coefficient (w _i ) 116 calculated by the weight calculation unit 110 and the plurality of feature amounts (f _i ) 11 calculated for each small block by the feature amount extraction unit 108.
4, the quantization control parameter calculation unit 111 calculates the quantization control parameter (z) 117 of the following equation.

[0022]

[Equation 3]

The quantization control parameter (z) 117 calculated by the quantum control parameter calculation unit 111 and the occupied state 118 of the encoded data buffer 106 are given to the quantization step calculation unit 112, and the quantization in the quantization unit 104 is performed. Step 119 is determined.

According to the configuration of FIG. 1, the weighting coefficient of each feature quantity is obtained from the distribution state of the feature quantity on the entire screen, and the effective feature weight is automatically quantized from the plurality of feature quantities. It can be reflected in the control.

FIG. 2 shows a block diagram of an embodiment of the present invention corresponding to the conventional block diagram of FIG. In the present embodiment, three types of feature amounts (intra-frame luminance signal variance, prediction error signal variance, and magnitude of difference vector between motion vector of small block and motion vector of adjacent small block) are used. In addition, weighting is performed based on the standard deviation obtained from the distribution state of each feature amount in the entire image to be encoded.

First, the image signal input from the input terminal 201 is divided by the small block dividing unit 202 into N × N small blocks. From the image signal 203 of this small block and the image of one frame before stored in the frame memory 214, the motion vector 2 is detected in the motion vector detecting unit 204.
05 is obtained, and based on this motion vector 205,
The motion compensation unit 206 performs motion compensation, and the subtractor 2
After the motion compensation prediction error signal 208 is obtained at 07,
The discrete cosine transform unit 209 performs discrete cosine transform on the motion compensation prediction error signal 208.

On the other hand, each feature amount is calculated as follows. First, the luminance variance calculator 215 calculates the luminance variance f ₁ within a small block from the image signal 203 divided into small blocks by the following equation.

[0028]

[Equation 4]

Further, from the motion vector 205 detected by the motion vector detection unit 104, the difference vector calculation unit 21
In step 6, the magnitude f ₂ of the difference vector is calculated by the following equation. In Equation 5, a represents a motion vector of a small block to be encoded, and b _i (i = 1, 2, ..., 8) represents motion vectors of eight small blocks vertically adjacent to each other and diagonally.

[0030]

[Equation 5]

Further, from the prediction error signal 208 output from the subtractor 207, the prediction error signal variance calculator 217 calculates the variance f ₃ of the prediction error signal in the small block by the following equation.

[0032]

[Equation 6]

Next, for each of the above three types of feature quantities f ₁ , f ₂ and f ₃ of the frame to be encoded, the feature quantity standard deviation calculation unit 218 makes each feature quantity for the entire frame. Then, the standard deviation of 223 is calculated and set to 223 as m ₁ , m ₂ , and m ₃ . From the calculated m ₁ , m ₂ , and m ₃ , the weight calculation unit 2
At 19, the weighting factors w ₁ , w ₂ , w ₃ for each feature amount are calculated by the following equation.

[0034]

[Equation 7]

From the weight coefficient w _i obtained by the weight calculation unit 219 and each feature amount f _i (i = 1, 2, 3), one feature amount (z) 225 is calculated by the quantization control parameter calculation unit 220. Obtained by an equation and used as the quantization control parameter.

[0036]

[Equation 8]

In the quantization step calculation unit 221, the quantization of the quantization unit 210 is performed based on the quantization control parameter (z) 225 obtained by the quantization control parameter calculation unit 220 and the buffer memory occupation amount 226 of the buffer 230. Step 227 is set. This set quantization step 2
27, the output discrete cosine transform coefficient 228 of the discrete cosine transform unit 209 is quantized by the quantizer 210,
After variable length coding is performed by the variable length coding unit 229, the variable length coding unit 229 inputs the variable length coding unit 229 into the buffer memory 230 and outputs the coded data output terminal 231 at a constant bit rate.

The discrete cosine transform coefficient quantized by the quantizer 110 is inversely quantized by the inverse quantizer 211, inverse discrete cosine transformed by the inverse discrete cosine transform unit 212, and then the motion compensator 106. Is added to the motion-compensated data of the previous frame output by the adder 213 and stored in the frame memory 214 for motion-compensated prediction of the next frame.

As described above, in the embodiment of FIG. 2, the number of feature amounts is 3, the feature amount is luminance dispersion, the motion vector difference vector,
Although the variance of the prediction error signal is used, the present invention is not limited to this. Furthermore, although the standard deviation of each feature amount in the image to be encoded is used as the feature of the entire image, other feature amounts obtained from the distribution state may also be used.
Further, the method for obtaining the weighting coefficient and the method for calculating the quantization control parameter can be arbitrarily determined.

[0040]

As described above, according to the present invention,
Quantization control can be adaptively performed on local features of the image signal. That is, in a region where deterioration is likely to be noticed, the deterioration can be made inconspicuous by performing fine quantization, and in a region where deterioration is not noticeable, the amount of information can be reduced by performing coarse quantization. Therefore, it is effective as a means for improving the image quality under a certain amount of information, and visual deterioration can be reduced.

[Brief description of drawings]

FIG. 1 is a basic configuration diagram of an image signal encoding device of the present invention.

FIG. 2 is a configuration diagram of an embodiment of the present invention when the number of feature amounts is three.

FIG. 3 is a configuration diagram of a conventional image signal encoding device.

FIG. 4 is a diagram showing an example of a region having uniform movement and a region having random movement.

FIG. 5 is a diagram showing an example of a region in which the motion is uniform and the intraframe variance is different.

[Explanation of symbols]

 101 input terminal 102 small block division unit 103 orthogonal transformation unit 104 quantization unit 105 code allocation unit 106 buffer 107 output terminal 108 feature amount extraction unit 109 feature amount distribution state calculation unit 110 weight calculation unit 111 quantization control parameter calculation unit 112 quantum Step calculation section

Claims (2)

[Claims] 1. An apparatus for encoding a digital image signal such as a television signal, means for dividing an image signal of a current frame into small blocks, means for obtaining a plurality of feature amounts for each small block, and one frame. A means for obtaining a distribution state for each feature amount across, a means for weighting each feature amount according to the obtained distribution state for each feature amount, and a plurality of feature amounts obtained for each small block Means for obtaining one new feature amount by performing weighting with the calculated weight.
An image signal encoding device, characterized in that the quantization characteristic of a small block to be encoded is switched using one feature amount. 2. The image signal coding apparatus according to claim 1, wherein the feature quantity includes an intra-frame luminance signal variance, a prediction error signal variance, and a difference vector between a motion vector of a small block and a motion vector of an adjacent small block. An image signal encoding device characterized by using a size.