JP3769770B2 - Quantizer and quantization method - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to a quantizer and a quantization method used for transmission or recording of a television signal or the like.
[0002]
[Prior art]
Conventionally, in order to transmit or record an image such as a television signal, the image has been encoded for compression. FIG. 3 shows a block diagram of a dynamic range adaptive encoding (hereinafter referred to as ADRC) apparatus as an example of the encoding. For example, data of 8 pixels × 8 lines (hereinafter referred to as (8 × 8) block) is supplied from the input terminal 21, and the input data of the (8 × 8) block is the maximum value detection circuit 22. To the minimum value detection circuit 26 and the subtractor 28. The maximum value detection circuit 22 detects the level at which the maximum value MAX of the pixels in the (8 × 8) block is detected and supplies the level to the subtractor 23. The minimum value detection circuit 26 determines the minimum value MIN of the pixels in the block. A level is detected. The detected minimum value MIN is supplied to the subtracters 23 and 28 and further taken out from the output terminal 27.
[0003]
The subtractor 23 generates a dynamic range DR by subtracting the minimum value MIN from the maximum value MAX, and the dynamic range DR is supplied to the step width calculation circuit 24 and taken out from the output terminal 25. In the step width calculation circuit 24, a desired step width is calculated from the supplied dynamic range DR, and the calculated step width is supplied to the quantization circuit 29. In the subtractor 28, the minimum value MIN is subtracted from the input data and normalized. The normalized value is quantized in the quantization circuit 29 based on the supplied step width, and the quantized value q is taken out via the output terminal 14.
[0004]
[Problems to be solved by the invention]
In the above-described conventional ADRC apparatus, for example, when input data is concentrated on a specific level such as an empty background, if the representative value when the input data is decoded differs from the specific level, there is an area. A problem that appeared as offset distortion occurred. For example, when dark luminance such as a shadow exists in one of two adjacent blocks including both the background sky, the dynamic ranges of the two blocks are greatly different, so that the distortions of the two blocks are different. Deterioration occurs so that the boundary of the visible.
[0005]
Accordingly, an object of the present invention is to provide a quantizer capable of encoding input data so that when the input data is concentrated at a specific level, pixels at that specific level are decoded at that level. And providing a quantization method.
[0006]
[Means for Solving the Problems]
The invention according to claim 1 forms a block composed of a plurality of pixels that are spatially or spatially and temporally close, detects a dynamic range that is a difference between the maximum value and the minimum value of the block, In a quantizer that normalizes a plurality of pixel values by a minimum value and quantizes the normalized pixel values with a step width adapted to the dynamic range,
The frequency of a plurality of pixels values in the block is measured, a level detecting means for detecting a level with 1/4 or more of the pixels included in the block at the maximum frequency as a specific level,
A correction amount that detects a difference between a detected specific level and a representative value that is a decoded value among values in a plurality of step widths, and calculates a difference that has the smallest absolute value among the differences as a correction amount Generating means;
Correction means for shifting the minimum value in the block by adding the correction amount to the minimum value;
It is a quantizer characterized by normalizing a plurality of pixel values in a block by a minimum value after shifting.
[0008]
[Action]
As for the input data, a specific level is detected for each block, the correction amount and the minimum value MIN are added so that the specific level becomes a representative value, and the input data is normalized based on the minimum value MIN ′ generated as a result. By quantizing the normalized input data, the dynamic range DR, the minimum value MIN ′, and the quantized value q are transmitted.
[0009]
【Example】
Embodiments of the present invention will be described below with reference to the drawings. FIG. 1 shows a block diagram of a quantizer according to an embodiment of the present invention. Input data divided into (8 × 8) blocks is supplied from the input terminal indicated by 1, and the input data is supplied to the level distribution measurement circuit 2, maximum value detection circuit 5, minimum value detection circuit 9 and subtractor 12. The The level distribution measurement circuit 2 counts the frequency for each level in the block, and the specific level detection circuit 3 sets 64 pixels in the (8 × 8) block to one level, for example, 1/4 or more. It is detected whether or not the pixels are concentrated.
[0010]
When 1/4 or more pixels are concentrated on one level, this level is set as a specific level, and the detected specific level is supplied to the correction amount calculation circuit 4. In the maximum value detection circuit 5 to which the input data is supplied, the maximum value MAX of the input data is detected for each block, and the detected maximum value MAX is supplied to the subtracter 6. In the subtractor 6, the minimum value MIN detected by the minimum value detection circuit 9 is subtracted from the maximum value MAX to generate a dynamic range DR. The dynamic range DR is supplied to the step width calculation circuit 7 and taken out from the output terminal 8.
[0011]
The step width calculation circuit 7 calculates a step width from the supplied dynamic range DR, the calculated step width is supplied to the quantization circuit 13, and a plurality of representative values corresponding to the calculated step width are obtained. It is supplied from the step width calculation circuit 7 to the correction amount calculation circuit 4. In the correction amount calculation circuit 4, the specific level supplied from the specific level detection circuit 3 is compared with a plurality of representative values supplied from the step width calculation circuit 7, and the correction amount Δ that minimizes the error is added to the adder 10. Supplied to. In the adder 10, the minimum value MIN and the correction amount Δ are added, and the addition result is supplied to the subtracter 12 as the minimum value MIN ′ and taken out from the output terminal 11.
[0012]
In the subtractor 12 , the input data is normalized based on the minimum value MIN ′, and the normalized input data is quantized in the quantization circuit 13 based on the supplied step width. The quantization value q is taken out from the output terminal 14. Further, in the level distribution measurement circuit 2, when 1/4 or more pixels are not concentrated on one level, the same processing as the conventional ADRC shown in FIG. 3 is performed.
[0013]
Here, an operation example of the present invention will be described with reference to FIG. As shown in FIG. 2A, the level distribution measurement circuit 2 measures the frequency of one block of input data for each level, and the specific level detection circuit 3 reduces the frequency for each level to ¼. A specific level where the above pixels are concentrated is detected. In the correction amount calculation circuit 4, a plurality of representative values based on the dynamic range DR are supplied from the step width calculation circuit 7, and the absolute value of the difference between the representative value and the specific level is obtained. The minimum value of the absolute values of these differences is detected, and this minimum difference is supplied to the adder 10 as the correction amount Δ. In the adder 10, the detected minimum value MIN and the correction amount Δ are added.
[0014]
That is, as shown in FIG. 2B, the minimum value MIN is shifted by the correction amount Δ. On the other hand, since the dynamic range DR is not changed, the minimum value MIN and the maximum value MAX are shifted by the correction amount Δ. As shown in FIG. 2B, the level distribution of the data after the shift is changed so that the specific level at which 1/4 or more pixels in the block are concentrated and the representative value are the same.
[0015]
In this embodiment, a method of shifting both the minimum value MIN and the maximum value MAX by the correction amount Δ is used in order to make the specific level and the representative value the same without changing the dynamic range DR. It is also possible to use a technique in which the specific level and the representative value are the same by changing the dynamic range DR by correcting the minimum value MIN and / or the maximum value MAX.
[0016]
Furthermore, although the above-mentioned embodiment is an example of a two-dimensional block of 8 pixels × 8 lines, the present invention can also be applied to a three-dimensional block using pixels in the time direction.
[0017]
In addition, the present invention is not limited to ADRC but can be applied to requantization for simply reducing the number of bits.
[0018]
【The invention's effect】
According to the invention, the quantizer of the image signal, when the input data are concentrated to a specific level, since the level is decoded as it is, is possible to remove the offset strain having an area It becomes possible.
[Brief description of the drawings]
FIG. 1 is a block diagram of an embodiment of a quantizer of the present invention.
FIG. 2 is a schematic diagram used to describe a specific level detection circuit according to the present invention.
FIG. 3 is a block diagram of an example of a conventional quantizer.
[Explanation of symbols]
2 level distribution measurement circuit 3 specific level detection circuit 4 correction amount calculation circuit 5 maximum value detection circuit 7 step width calculation circuit 9 minimum value detection circuit 13 quantization circuit

Claims (1)

A block composed of a plurality of pixels that are spatially or spatially and temporally close is formed, a dynamic range that is a difference between the maximum value and the minimum value of the block is detected, and a plurality of pixel values in the block are set to the minimum In a quantizer that normalizes by a value and quantizes the normalized pixel value with a step width adapted to the dynamic range,
The frequency of a plurality of pixels values in the block is measured, a level detecting means for detecting a level with 1/4 or more of the pixels included in the block at the maximum frequency as a specific level,
The difference between the detected specific level and the representative value that is the decoded value among the values in the plurality of step widths is detected, and the difference having the smallest absolute value among the differences is calculated as the correction amount. Correction amount generating means to perform,
Correction means for shifting the minimum value in the block by adding the correction amount to the minimum value;
A quantizer characterized by normalizing a plurality of pixel values in the block by a minimum value after shifting.

Images