JPH01256285A - Picture signal band compression system - Google Patents

Abstract

PURPOSE:To reduce deterioration in picture quality due to complicated combination of limited kinds of modes with a larger picture quality difference among adjacent blocks in spatial or timewise directions by applying filtering to the mode decided once in the spatial or timewise directions so as to correct the mode. CONSTITUTION:An interpolation error to each mode is calculated and the total sum in a block of the absolute value of the interpolation error, for example, is calculated by using the said error and the result is used as a distortion for each block. Then the mode is decided by using the distortion for each block. After the determination in this way, filtering is applied to the mode decided once in the spatial or timewise direction to correct the mode. Thus, the possibility of the assignment of the same mode in the spatial or timewise direction is enhanced. Then the deterioration in picture quality due to a random change of the mode in the timewise direction caused by noise depending on complicated combination of limited kinds of modes with a larger picture quality difference for each block timewise directions is reduced.

Description

[Detailed Description of the Invention] [Field of Industrial Application] The present invention relates to an image signal band compression method,
In particular, it relates to an image signal band compression method that can be applied to image signal transmission or recording.

(Prior art) Conventionally, regarding band compression technology for image signals, a technology called TAT method is known, which is reported in the Institute of Electronics and Communication Engineers Communication Systems Study Group Material C384-7r "Time axis conversion band compression method for high-definition television signals". ing.

The TAT method first extracts basic pixels representing the basic structure of an image by coarse subsampling, and then selects additional pixels necessary to represent fineness from among the remaining pixels.

This will be explained with reference to FIGS. 7 and 8.

In FIG. 7, which is an explanatory diagram showing a sampling pattern of the TAT method, symbols p+ and pg indicate a sampling point and a thinning point, respectively, and an arrow M indicates an interpolation method.

The above processing in the TAT method divides the image into small blocks, and divides the image into small blocks according to the fineness of the image within the block.
This corresponds to changing the sampling pattern as shown in Figures (a) and (b).

That is, a plurality of modes are set with one sampling pattern as one mode, the mode is switched and assigned to each block according to the fineness of the image, and pixels are thinned out and band compression is performed according to the assigned mode. By doing this, pixels are not thinned out in fine parts of the image,
It is possible to reduce the interpolation error as a whole by thinning out pixels in flat parts of the image. In determining the mode in the TAT method, the number of modes is as shown in FIG. 7(a).

In the case of the two types shown in (b), the seventh
Using the histogram of the distortion ilS for each block, which is the intra-block summation of the absolute value of the interpolation error for mode 2 in FIG. 2(b), the mode is determined using the value Th of S as the boundary. That is, for blocks exceeding Th, mode 1 is
The following blocks are assigned mode 2.

[Problem to be solved by the invention]

However, band compression using such a method still has the following problems in terms of image quality.

That is, in the conventional TAT method, the mode for each block is determined by looking at the distortion amount value for each block, which is the intra-block summation of the absolute value of the interpolation error, and using the dark value as the boundary, as described above. There is. Therefore, the amount of distortion for each block is the threshold T
In the image area near h, limited types of modes with large differences in image quality are complicated for each block, and this appears as visual image quality deterioration. Furthermore, in this area, even a slight change in the amount of distortion in the time direction due to the influence of noise etc. may cause the mode of blocks at the same position to change randomly from screen to screen, which also causes visual image quality deterioration. It appears.

An object of the present invention is to solve the above-mentioned problems by filtering the modes determined using the amount of distortion for each block in the spatial or temporal direction and correcting the modes where the modes are complicated. The object of the present invention is to provide a relaxed image signal band compression method.

[Means to solve the problem]

The image signal band compression method of the present invention divides an image into blocks of a predetermined size, determines a mode indicating the pixel thinning and interpolation method for each block on the compression side, and processes the image according to the mode. In an image band compression method in which pixels are thinned out and restored by interpolating the thinned out pixels on the restoration side, when determining the mode, an interpolation error for each mode is calculated for each pixel, and at least the interpolation error is used. The amount of distortion for each block is calculated using the amount of distortion for each block, the mode determined using the amount of distortion for each block is filtered in the spatial or temporal direction, this is set as a correction mode, and the pixels are thinned out according to the correction mode. It is characterized by compression.

[Effect]

One of the features of the present invention is the mode determination and correction method in image band compression processing, in which pixels are thinned out on the transmitting side in the case of transmission, and the thinned out pixels are interpolated on the receiving side to restore the image. be.

In the present invention, first, the interpolation error for each mode is calculated, and this is used to calculate, for example, the intra-block total of the absolute value of the interpolation error, and this is used as the distortion amount for each block. Furthermore, the mode is determined using the amount of distortion for each block.

After determining in this way, the determined mode is modified by applying a filter in the spatial or temporal direction.

Specifically, mode modification can be performed by further filtering the mode using modes of a plurality of blocks located around the screen or at the same position but at different times. As described above, - Apply a filter to set this as a modification mode, and according to this modification mode, pixels are thinned out from the image signal and the band is compressed.

By correcting the mode by applying a filter in the spatial or temporal direction to the mode determined using the distortion amount for each block, the possibility that the same mode will be assigned in the spatial or temporal direction increases. It is possible to reduce image quality deterioration caused by limited types of modes with large differences in image quality depending on the time, which are caused by random changes in the mode in the temporal direction due to spatially complex noise.

〔Example〕

Next, the present invention will be explained with reference to the drawings.

1 to 6 show embodiments of the present invention, and are respectively a block diagram showing an outline of the basic part, a diagram showing an example of its mode determining circuit, a diagram showing a spatial region used in the filter, and a diagram showing the filter. FIG. 7 is a diagram illustrating the effect of , another example of the type of mode to which the present invention can be applied (another example different from the case of FIG. 7), and a diagram illustrating the effect of the filter.

First, an embodiment of the present invention will be described with reference to FIGS. 1 to 4, and FIGS. 7 and 8.

In this embodiment, as an example, a television signal is input, a field is divided into blocks each having a size of 4 pixels x 4 pixels, a mode is determined for each block for each field, and the number of pixels is reduced to 1 of the total. The case of thinning out to /2 is shown. The mode in this embodiment is shown in FIG. 7(a) as the second side.

Mode 1 shown in (b). There are two types: Mode 2.

The compression ratio of each mode is 1 for mode 1 and 1/4 for mode 2, and mode 2 uses intra-field interpolation.

That is, in this embodiment, two types of modes (a) and (b) in FIG. 7 are assigned to each block, and bandwidth compression is performed by thinning out corresponding to each block. I will explain about it.

FIG. 1 is a block diagram showing the basic structure of a transmitting section of a television signal band compression system according to an embodiment of the present invention.

As shown in FIG. 1, the transmitter includes a mode 2 thinning interpolation circuit 1 to which an input image signal (original signal) 50 is supplied from an input terminal 100, an interpolation signal 51 from this interpolation circuit 1, and an input image signal 50. , a distortion calculation circuit 3 to which the difference signal 52 is supplied, and a mode determination circuit 4 to which the output distortion signal 53 is applied. Details of this mode determining circuit 4 will be described later.

Furthermore, the transmitter includes a mode correction circuit 5, a delay circuit 6,
The mode signal 54 from the mode determining circuit 4 is supplied to the mode modification circuit 5, and the modified mode signal 55 obtained by the mode modification circuit 5 is sent to the output terminal 102. supplied to 7,
The delayed input image signal 56 obtained through the delay circuit 6 is thinned out, and the thinned out image signal 57 is sent to the output terminal 101 as an output.

Now, to explain with reference to Figure 2 and below, input terminal 1
When the image signal 50 is added to 00, the mode 2 thinning and interpolation circuit 1 performs mode 2 thinning and interpolation on the image signal 50 input from the input terminal 100, and the interpolated signal 51 is input to the difference circuit 2. Then, the difference between the interpolated signal 51 and the original signal, that is, the image signal 50 is calculated.

The difference signal 52 is input to the distortion amount calculation circuit 3, and for example, the sum of the absolute values of the difference signal 52 within a block is calculated and this is taken as the distortion amount D2 for each block.

The amount of distortion for each block is output as a distortion amount signal 53.

The distortion amount signal 53 is input to the mode determining circuit 4, and the mode is determined in the same manner as in the conventional TAT method, as shown in FIG.

Here, the mode determining circuit 4 includes a histogram creating circuit 8, a dark value determining circuit 9, a mode assigning circuit 10, and a delay circuit 11, as shown in FIG. In FIG. 2, first, when the distortion amount signal 53 is applied, it is input to the histogram creation circuit 8, and a histogram of the distortion amount D2 is created in the same manner as shown in FIG. In the threshold determination circuit 9, the contents of the histogram created in the histogram creation circuit 8 are read out one after another from the larger D2 and added up, and
The value of D2 when it becomes 1/3 or more is set as a threshold Th, and is output as a threshold signal 59. Furthermore, the distortion amount signal 53 is
Until the i value Th is determined by the dark value determination circuit 9, it is delayed by the delay circuit 11. The mode assignment circuit 10 uses the delayed distortion amount signal 58 and the threshold signal 59 to determine the mode for all blocks. That is, for each block, mode 1 is assigned to blocks whose distortion amount D2 is greater than or equal to Th, and mode 2 is assigned to blocks whose distortion amount D2 is smaller than Th, and the mode signal 54 is output.

Returning to FIG. 1, the mode signal 54 determined by the mode determining circuit 4 is input to the mode modifying circuit 5, where the mode signal is filtered in the spatial or temporal direction. For example, as shown in FIG. 3, looking at an area of 3 blocks x 3 blocks centered around the block of interest labeled 30, among the 8 blocks (block mill block h) other than the block of interest 30, for example, If seven or more blocks are in the same mode (mode i) and the block of interest 30 is in another mode (mode j: 1≠j), the mode of the block of interest 30 is replaced with mode i, It is output as a correction mode signal 55.

This situation is shown in FIG.

As described above, in this method, first, the interpolation error for each mode is calculated, and this is used to calculate, for example, the intra-block sum of the absolute values of the interpolation errors, and this is used as the distortion amount for each block. Furthermore, after determining the mode using the distortion amount for each block in the same manner as the conventional TAT method,
Furthermore, the mode signal is filtered using the modes of multiple blocks located around the screen or at the same position but at different times, and this is set as a correction mode, and as described later, the image is imaged according to this correction mode. Thin out pixels from the signal to compress the band. As a filter, for example, as shown in Figure 4, if the mode of the peripheral part of the block of interest is the same and only the block of interest is in a different mode from the peripheral part, the mode of the block of interest is corrected to the mode of the peripheral part. By using a filter that has the ability to do this, it is less likely that limited types of modes with large differences in image quality will be complicated. Furthermore, by applying a filter that suppresses mode changes in the time direction, random changes in the mode in the time direction due to the influence of noise etc. can be suppressed.

In this way, by modifying the mode determined using the distortion amount for each block by applying a filter in the spatial or temporal direction, the possibility that the same mode will be assigned in the spatial or temporal direction increases. . Therefore, deterioration in image quality due to spatially complex and intricate limited types of modes with large differences in image quality for each block, or random changes in modes in the time direction due to noise, etc., is reduced.

On the other hand, in the system of the delay circuit 6 and thinning circuit 7 shown in FIG. 5o is delayed by a delay circuit 6. Then, the pixels of the delayed input image signal 56 are thinned out in accordance with the correction mode signal 55 in the thinning circuit 7, and the thinned out image signal 57 and the correction mode signal 55 corrected in the mode correction circuit 5 are respectively output to the output terminal 101. .102 to the receiving side.

Below, a mode determination method based on this embodiment will be further explained.

In the embodiment shown in FIG. 1, the mode is determined for each field, but the mode may be determined for a plurality of fields at once.

Furthermore, before the mode 2 thinning interpolation circuit 1 and thinning circuit 7 in FIG. 1 perform thinning, the signal band may be limited in order to prevent aliasing distortion in correspondence with mode 2.

Furthermore, as a method for determining the threshold value Th in the threshold value determination circuit 9 of FIG.
You can start with the smaller one. However, in that case,
Let Th be the value of D2 when the addition result exceeds 2/3 of the total number of blocks.

In addition, as a filter in the mode correction circuit 5 of FIG.
In the block area, among the eight blocks other than the block of interest (block mill block h), four blocks, block 1 block d, block f, and block h, are in the same mode (mode i), and the block of interest is If the block is in another mode (mode j; i≠j), the mode of the block of interest may be replaced with mode i.

Furthermore, a number corresponding to the mode is assigned to each block (for example, 11 for a mode 1 block, 2 for a mode 2 block), a low-pass filter is applied to the assigned number, and the filter output is A similar effect can be obtained by allocating a new mode based on the value and using this as a modified mode signal.

Also, for example, first filter only the blocks in mode 1 and count the number of blocks that have changed from mode 1 to mode 2, then filter the blocks in mode 2 to count the blocks that have changed from mode 1 to mode 1. By replacing mode 1 with mode 2 by the number of blocks replaced with mode 2, changes in the compression ratio due to mode modification can be suppressed. In addition, instead of focusing on each block individually and filtering them individually, for example, in an area where only mode 1 is concentrated, n blocks are grouped together as mode 2, and n blocks are in mode 2. If an isolated block exists in a mode 1 block, all n mode 2 blocks may be replaced with mode 1. Furthermore, as a filter, for example, a filter may be applied in the temporal direction in exactly the same manner as in the spatial direction, using modes of blocks at the same position but at different times.

The types of modes used are shown in FIG. 7(a).

It may be different from (b), for example, Fig. 5(a),
Mode 1 shown in (b) and (c). Three types, mode 2 and mode 3, may be used. The compression ratio of each mode is 1 for mode 1 in Figure 5(a), 1/2 for mode 2 in Figure 5(b), 1/4 for mode 3 in Figure 5(C), and 1/4 for mode 2 in Figure 5(C).
．． Mode 3 uses the inside field. FIG. 6 shows the effect of mode correction in this case.

According to this embodiment, the signal band of the television signal is halved.
The bandwidth of the transmission path required for transmission is 1/2.
This enables an image signal band compression method with even better image quality. Note that the compression rate of the image signal can be varied by controlling the compression rate or ThO value of each mode.

〔Effect of the invention〕

As described above, according to the present invention, by correcting the mode by applying a filter in the spatial or temporal direction to the mode determined by − degrees, image quality differences between blocks adjacent in the spatial or temporal direction are obtained. It is possible to reduce image quality deterioration caused by a complicated combination of limited types of modes with large numbers. Therefore, it is possible to perform image band compression with visually higher quality than the conventional TAT method.

[Brief explanation of the drawing]

FIG. 1 is a schematic block diagram of the basic part of a transmitter for explaining an image signal band compression method according to an embodiment of the present invention, and FIG. 2 shows an example of a mode determination circuit in the embodiment of FIG. 3 is a diagram showing the spatial area used for the filter in the embodiment of FIG. 1, FIG. 4 is a diagram showing the effect of the filter in the embodiment of FIG. 1, and FIG. 5 is a diagram showing the types of modes, etc. 6 is a diagram showing the effect of the filter in that case, FIG. 7 is a diagram showing the sampling pattern of the TAT method, and FIG. 8 is a diagram showing the mode determination method in the TAT method. 1...Mode 2 decimation interpolation circuit 2...Difference circuit 3...Distortion amount calculation circuit 4...Mode determination circuit 5...Mode correction circuit 6... ... Delay circuit 7 ... Thinning circuit 8 ... Histogram creation circuit 9 ... Threshold determination circuit IO ... Mode assignment circuit 11 ... Delay circuit agent Patent Attorney Yoshiyuki Iwasa 3 Zorotsuk Figure 3 × × × ×
Mode 2(a)
(1)) Figure 7

Claims (1)

[Claims] (1) Divide the image into blocks of a predetermined size, determine a mode indicating the pixel thinning and interpolation method for each block on the compression side, thin out pixels on the compression side according to that mode, and then on the restoration side In an image band compression method that restores thinned pixels by interpolation, when determining the mode, an interpolation error for each mode is calculated for each pixel, and at least the amount of distortion for each block is calculated using the interpolation error. The method is characterized in that the mode calculated and determined using the distortion amount for each block is filtered in the spatial or temporal direction to set it as a correction mode, and pixels are thinned out according to the correction mode and the band is compressed. Image signal band compression method.