JPS6232785A - Adaptive vector quantization system - Google Patents

Abstract

PURPOSE:To remarkably reduce the volume of calculation for the selection of reproducing vector by detecting the luminance change of an input signal by block unit and selecting the reproducing vector corresponding to the volume of the luminance change. CONSTITUTION:An inputted block is inputted to a vector quantizer 20 and a luminance change measuring and rank deciding part 10. At the luminance change measuring and rank deciding part 10, the luminance change of the inputted block is measured and a rank is decided corresponding to the volume of the luminance change, for example, 1, 2,...n from the smallest, and a rank information is sent to the vector quantizer 20, and also, is transmitted to a transmission path. At the vector quantizer 20, either of vector quantizers 1-n is selected and quantized corresponding to the rank information and the information quantized, and code information is sent to the transmission path. Thereby, the calculation volume for a selection at the vector quantizer can be reduced.

Description

[Detailed Description of the Invention] [Already Required] When vector quantizing an image signal, a change in brightness of the human input signal in units of blocks is detected, and a reproduction method is prepared correspondingly depending on the magnitude of the change in brightness. By selecting a reproduction vector from a codebook that has vectors,
This reduces the amount of calculation required to select a reproduction vector.

[Industrial application field]

The present invention relates to improvements in vector quantization methods for image signals.

Vector quantization has recently come into use as a coding method for image signals.

As shown in an example in Fig. 2, the screen is divided into 4 x 4 pixel blocks (16-dimensional vectors), and various reproduction vectors of various luminance information that are expected to occur in these blocks are used. Create a code, add a code to it, prepare a codebook that collects these, and when transmitting an image signal, select the reproduction vector closest to the luminance information of this block in the codebook for each block. In this method, the degree of band compression is increased by transmitting the code of this reproduction vector.

When performing this vector quantization, it is desirable that the amount of calculation for selecting reproduction vectors is small.

[Prior art and problems to be solved by the invention] In the conventional vector quantization method, when transmitting an image signal,
For each block, the reproduction vector closest to the species information of this block in the codebook is selected, and the code of this reproduction vector is transmitted.

In this case, in order to reproduce a beautiful picture, a small number of reproduction vectors can be used for flat areas with little change in brightness (areas with small changes in brightness), but for edge parts where there are many changes in brightness such as the contours of the picture, Many reproduction vectors are required. Therefore, in order to reproduce a beautiful picture, a large number of reproduction vectors corresponding to both the flat part and the edge part are required, and the selection is made from among these, which solves the problem of increasing the amount of calculation required to select the reproduction vector. be.

Note that the change in brightness P includes the case shown in the following equation when the brightness level of a pixel is defined as Xi.

P=1/ng(Xi-Xl)2, P=1/nff1 IXi -Xl,I [Means for solving the problem] The above problem is based on detecting the change in brightness of the input signal for each program. Then, a reproduction vector is selected from a codebook with reproduction vectors prepared for each magnitude of change in brightness, and rank information on the magnitude of change in brightness and a code of the reproduction vector are transmitted. This problem is solved by the adaptive vector quantization method of the present invention.

[Effect]

According to the present invention, a codebook with a small number of reproduction vectors is given to a flat part where the luminance changes are small, and a codebook with a corresponding large number of reproduction vectors is given to an edge part where the luminance variation is large. As described above, codebooks corresponding to the magnitude of the change in brightness are provided, and reproduction vectors are selected from each codebook according to the magnitude of the change in brightness, so the amount of calculation for selection is as follows. Very little on the flat parts, and very little on the edges.

Since the amount of calculation is also smaller than that of the conventional method, the overall amount of calculation can be greatly reduced.

〔Example〕

FIG. 1 is a block diagram of an embodiment of the invention.

In the figure, 10 is a luminance change measurement and rank determination unit, 20.1
,2. n is a vector quantizer, 1 is a vector quantizer with few reproduction vectors corresponding to the small power part,
n is a vector quantizer with a large number of reproduction vectors corresponding to a portion with a large change in brightness, and 2 is a vector quantizer with an intermediate number of reproduction vectors corresponding to an intermediate portion of a change in brightness.

The input blocks are input to the vector quantizer 20 and the brightness change measurement and rank determination unit IO.

The brightness change measurement and rank determination unit 10 measures the brightness change of the input block, and determines the rank according to the magnitude of the brightness change, for example, 1.2...n from the side with the smallest brightness change. , this rank information is transmitted to the vector quantizer 10
It is also sent to the transmission line.

The vector quantizer 20 selects one of the vector quantizers 1, 2, .

In this way, the number of reproduced vectors in vector quantizer 1 is small, so the amount of calculation is very small, and as the vector quantizer 2...n increases, the number of reproduced vectors increases, so the amount of calculation increases. , vector quantizer n requires less calculation for selection than the conventional one, so the overall calculation amount can be significantly reduced.

〔Effect of the invention〕

As described in detail above, according to the present invention, when vector quantizing an image signal, the amount of calculation for selecting a reproduction vector can be extremely reduced.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of the present invention, and FIG. 2 is a diagram of dividing an example screen into blocks in the case of vector quantization. In the figure, l,'l,n, 20 is a vector quantizer, and 10 is a luminance change measurement and rank determination unit. bet) rate t-+-1 in other cases h? Phantom 17
Minute 1 figure 20 to 0-77

Claims (1)

[Claims] When vector quantizing an image signal, changes in brightness in blocks of the input signal are detected, and a codebook is created that has reproduction vectors prepared in accordance with the magnitude of the change in brightness. An adaptive vector quantization method is characterized in that a reproduction vector is selected from among them, and rank information on the magnitude of change in brightness and a code of the reproduction vector are transmitted.