JPH04229790A - Transmitter for picture data - Google Patents

Abstract

PURPOSE:To obtain a picture with high picture quality when the picture is decoded by extracting a local feature of the picture for each block and generating a selection signal of plural sampling patterns based thereon. CONSTITUTION:An input picture signal from an input terminal 1 is converted into a digital picture element data by an A/D converter 2, fed to a block processing circuit 3 and divided into block data comprising plural picture element data. The block data from the circuit 3 is fed to a subsampling circuit 4. A subsampling signal from a sampling signal generating means 5 selected by a selection signal from a selection signal generating means 8 is fed to the circuit 4 and the block data from the circuit 3 is sub-sampled. On the other hand, the block data from the circuit 3 is fed to a horizontal correlation detection circuit 6 and a vertical correlation detection circuit 7 and values X, Y representing the degree of a change are fed from the circuits 6,7 to the circuit 8 and a selection signal SE is fed to the circuit 5.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image data transmission device capable of transmitting image data at a low data rate.

0002

[Prior Art] When transmitting digital image data,
Subsampling technology is well known as one of the effective means for reducing the transmission data rate.

In this method, for example, thinning processing is performed in units of pixel data, and the thinned out pixel data is restored by performing interpolation processing using adjacent pixel data in the thinning direction. Therefore, the quality of the restored image is good when the degree of change between pixel data in the thinning direction is small, that is, when the correlation between pixel data in the thinning direction is strong.

[0004] Therefore, the thinning pattern, that is, the subsampling pattern, is appropriately selected in consideration of the direction in which the correlation between the pixel data is strong. In many cases, only one type of sub-sampling pattern is selected.

[0005]

However, when image signals are digitized and transmitted, it is common practice to divide one screen's worth of image data into a plurality of blocks and perform processing on a block-by-block basis. Typically, one block includes n pieces of pixel data in the horizontal direction (n is an integer of 2 or more) and m lines, that is, m pieces of pixel data (m is an integer of 2 or more) in the vertical direction. Therefore, one block is n×m
consists of pixel data.

Subsampling processing is also performed on a block-by-block basis, but the features of each block, which are local features of the image, often differ from block to block. That is, the direction in which the correlation between pixel data is strong often differs from block to block. Therefore, as mentioned above, when subsampling processing is performed for each block using one type of subsampling pattern, the image quality of the restored image does not match the characteristics of each block, that is, the local characteristics of the image. sometimes deteriorated.

SUMMARY OF THE INVENTION In view of the above points, it is an object of the present invention to provide an image data transmission device that can perform subsampling according to local characteristics of an image and improve image quality.

[0008]

[Means for Solving the Problems] In order to achieve the above object, an image data transmission device according to the present invention has the following features:
Corresponding to the reference numerals in the embodiment of FIG. On the other hand, a subsampling means 4 performs subsampling with a thinning pattern selected from a plurality of different thinning patterns by a selection signal, a horizontal correlation detecting means 6 detecting a correlation in the horizontal direction of each of the block data, Vertical correlation detection means 7 for detecting vertical correlation of each block data
and a selection signal forming means 8 for forming a thinning pattern selection signal SE from the correlation detection output of the horizontal correlation detection means 6 and the correlation detection output of the vertical correlation detection means 7,
The image data from the sub-sampling means 8 and the selection signal SE from the selection signal forming means are used as transmission data.

[0009]

[Operation] The horizontal correlation detection means 6 and the vertical correlation detection means 7 detect image features for each block. In the selection signal forming means 8, a selection signal is formed from the correlation detection outputs of the horizontal correlation detection means 6 and the vertical correlation detection means 7,
Based on this selection signal, an appropriate pattern is selected as a thinning pattern by the sub-sampling means 4 according to the characteristics of the image of each block.

0010

DESCRIPTION OF THE PREFERRED EMBODIMENTS FIG. 1 is a block diagram of an embodiment of an image data transmission apparatus according to the present invention. The main parts of this example can also be realized by microcomputer software, in which case the parts represent the functions performed by the program software.

In FIG. 1, an input image signal is supplied to an A/D converter 2 through an input terminal 1, sampled at a predetermined sampling frequency, and each sampled value is converted into digital pixel data consisting of, for example, 8 bits. .

The digital pixel data from the A/D converter 2 is supplied to a blocking circuit 3, where it is divided into block data consisting of a plurality of pixel data. in this case,
Each block data includes a plurality of pixel data in the horizontal and vertical directions, and the image data for one screen is divided into four pixels in the horizontal direction, and The data is divided into four pixel data units (four lines) to form one block of 4×4=16 pixel data.

The block data from this blocking circuit 3 is supplied to a sub-sampling circuit 4. This subsampling circuit 4 is supplied with a subsampling signal from a subsampling signal forming means 5.

In this case, the subsampling signal forming means 5 can generate subsampling signals of a plurality of different subsampling patterns, that is, a plurality of thinning patterns. Then, a subsampling signal selectively selected from among the subsampling signals of the plurality of types of thinning patterns by a selection signal from a selection signal forming means 8, which will be described later, is supplied to the subsampling circuit 4. ing.

In this example, the subsampling signal forming means 5 includes first to third subsampling signal generating sections 51 and 52 that generate subsampling signals of three types of thinning patterns as shown in FIG. , 53. Note that in FIG. 3, ◯ marks indicate pixel data to be transmitted, and × marks indicate pixel data to be thinned out.

The first sub-sampling signal generating section 51 generates a sub-sampling signal in a so-called horizontal 1/2 thinning grid pattern, in which pixel data is thinned out every pixel data in the horizontal direction, as shown in FIG. 3A. .

The second subsampling signal generating section 52 also generates a subsampling signal in a so-called vertical 1/2 thinning grid pattern, in which pixel data is thinned out every pixel data in the vertical direction, as shown in FIG. 3B. Occur.

Furthermore, the third sub-sampling signal generating section 53 thins out the pixel data for each pixel data in both the horizontal and vertical directions as shown in FIG. 3C.
A sub-sampling signal having a so-called quincunx-shaped 1/2 thinning grid pattern is generated.

The subsampling signals from each of these subsampling signal generators 51, 52, and 53 are supplied to a selector 54, and the subsampling signals selected by the selector 54 as described later are used as subsampling signals. It is supplied to circuit 4. The selection signal of the selection section 54 is formed by the selection signal forming means 8 as follows. That is, the block data from the blocking circuit 3 is
The signal is supplied to the horizontal correlation detection circuit 6 and also to the vertical correlation detection circuit 7.

The horizontal correlation detection circuit 6 calculates the average value of the absolute values of the first-order differences in the horizontal direction X X=Σ|dxi|/Σi. Here, as shown in FIG. 2, dxi is the difference between the i-th pixel data and the i+1-th pixel data in the horizontal direction. The value X represents the degree of change in pixel data in the horizontal direction, and the smaller the value X, the smaller the change, indicating that the correlation in the horizontal direction is stronger.

The vertical correlation detection circuit 7 calculates the average value Y Y=Σ|dyi|/Σi of the absolute values of the first-order differences in the vertical direction. Here, dyi is the difference between the i-th pixel data and the i+1-th pixel data in the vertical direction, as shown in FIG. The value Y represents the degree of change in pixel data in the vertical direction, and the smaller the value Y, the smaller the change, indicating that the correlation in the vertical direction is stronger.

The value X thus determined by the horizontal correlation detection circuit 6 and the value Y determined by the vertical correlation detection circuit 7 are supplied to the selection signal forming means 8. This selection signal forming means 8
consists of first to third comparing sections 81 to 83, a threshold value generating section 84, and a selection signal forming section 85.

The first comparing section 81 is supplied with the values X and Y, the magnitudes of the two are compared, and the comparison output is supplied to the selection signal forming section 85 . Further, the value X is supplied to the second comparing section 82 and compared with the threshold value K from the threshold value generating section 84 , and the comparison output is supplied to the selection signal forming section 85 . Furthermore, the value Y is supplied to the third comparing section 83 and compared with the threshold value K from the threshold value generating section 84 , and the comparison output is supplied to the selection signal forming section 85 .

The selection signal forming section 85 generates a 2-bit selection signal SE based on the following conditions.

■If X<K and Y<K, and if X<Y, the selection signal SE="00" is set. At this time, a horizontal 1/2 thinning grid pattern is selected as the thinning pattern.

■If X<K and Y<K, and if X>Y, the selection signal SE is set to "01". At this time, a vertical 1/2 thinning grid pattern is selected as the thinning pattern.

■ If X>K and Y>K, and X<Y, the selection signal SE is set to "10". Also, X>K and Y>K
If X>Y, the selection signal SE="11" is set. Furthermore, if X=Y, the selection signal SE is set to either "10" or "11". At this time, a vertical 1/2 thinning grid pattern is selected as the thinning pattern.

The selection signal SE thus formed is supplied to the selection section 54 of the subsampling signal generation means 5, and selection control is performed as described above. Then, the subsampling signal of the thinning pattern selected by the selector 54 is supplied to the subsampling circuit 4, and the block data from the blocking circuit 3 is subsampled.

In this example, even if the subsampling patterns are different, the thinning rate is the same, so the output block data of the subsampling circuit 4 is composed of the same number of pixel data. The output block data of this sub-sampling circuit 4 is processed by a block encoding circuit 9.
After being further subjected to high-efficiency encoding processing and data compression, the image data is output to the output terminal 10 as output image data, and is used as data for transmission.

Since the number of constituent pixel data of each block subjected to the thinning process from the sub-sampling circuit 4 is always constant as described above, the block encoding process in the block encoding circuit 9 is easy. This block encoding circuit 9
As the block encoding process, for example, ADRC (Adaptive Dynamic
Range Coding; JP-A-61-14498
No. 9, JP-A-62-92620, JP-A-62
128621), an orthogonal transform encoding method such as DCT, a vector quantization method, and other encoding methods can be used.

Further, the selection signal SE from the selection signal forming means 8 is led out to the output terminal 11 and used as additional data together with the output image data as data for transmission. In this case, on the image data decoding side, the selection signal SE indicates the interpolation direction when restoring the thinned out pixel data by interpolation.

That is, on the image data decoding side, after performing block decoding processing, the selection signal SE is referred to, and if the selection signal SE is "00", the thinning direction of the transmitted image data is horizontal. Therefore, the thinned out pixel data is restored by interpolation processing using horizontally adjacent pixel data. For example, average value interpolation is performed using pixel data on both sides in the horizontal direction.

Furthermore, if the selection signal SE is "01",
Since the thinning direction of the transmitted image data is the vertical direction, the thinned out pixel data is restored by interpolation processing using vertically adjacent pixel data.

If the selection signal SE is "10", it means that the transmitted image data has been subsampled with a 5-mesh thinning grid pattern, but as described above, in this case, X<Y , and the horizontal correlation is stronger than the vertical correlation. Therefore, interpolation of the thinned out pixel data is performed using horizontally adjacent pixel data. Also, if the selection signal SE is "11", this also means that the transmitted image data is subsampled with a 5-mesh thinning grid pattern, but as described above, X>Y
, and the vertical correlation is stronger than the horizontal correlation. Therefore, interpolation of the thinned out pixel data is performed using vertically adjacent pixel data. Therefore, the pixel data is thinned out in both the horizontal and vertical directions.
Even in the case of the mesh-shaped thinning lattice pattern, interpolation processing can always be performed in the direction of strong correlation, leading to an improvement in the quality of the restored image.

As described above, local features of the image for each block are extracted, and selection signals for a plurality of sub-sampling patterns are formed based on the extracted features.
Subsampling is performed using a subsampling pattern suitable for the characteristics of each block. Therefore, when the image is restored, a high quality image can be obtained.

In the case of the above example, even in the case of the 5-mesh thinning lattice pattern in which pixel data is thinned out in both the horizontal and vertical directions, the selection signal SE is divided into two ways, and which of the horizontal direction and the vertical direction is correlated. Since the direction of interpolation can be indicated by the strength of the image, further improvement in image quality can be expected. Moreover, in this example, since the selection signal SE is 2 bits, it is originally possible to distinguish between 4 types, but
There are three types of sub-sampling patterns, and one option remains. Therefore, as in this example, even if two types of selection signals are used in the case of a 5-mesh thinning grid pattern, there is no need to add one additional bit, and the 2-bit selection signal SE is used. Can be used without waste.

In the above example, three types of sub-sampling patterns were used as sub-sampling patterns, but it is of course possible to prepare four or more types of patterns. Furthermore, it goes without saying that the thinning pattern is not limited to the above example, and various other patterns can be used.

Furthermore, in the above example, three types of sub-sampling patterns are prepared, all of which have the same thinning rate of 1/2, but a plurality of types of sub-sampling patterns with different thinning rates may be prepared.

[0039]

As explained above, according to the present invention, multiple types of sub-sampling patterns are prepared in advance, local features of the image for each block are extracted, and based on the local features of the image, multiple types of sub-sampling patterns are prepared. Since the selection signal for each type of sub-sampling pattern is formed, sub-sampling can be performed using a sub-sampling pattern suitable for the characteristics of each block. Therefore, when the image is restored, a high quality image can be obtained.

[Brief explanation of the drawing]

FIG. 1 is a block diagram of an embodiment of an image data transmission device according to the present invention.

FIG. 2 is a diagram showing an example of the configuration of one block.

FIG. 3 is a diagram illustrating an example of multiple sub-sampling patterns.

[Explanation of symbols]

3 Blocking circuit 4 Subsampling circuit 5 Subsampling signal forming means 6 Horizontal correlation detection means 7 Vertical correlation detection means 8 Selection signal forming section

Claims (1)

[Claims] 1. Blocking means for dividing image data into blocks each consisting of a plurality of pixel data in the horizontal and vertical directions, and dividing the block data from the blocking means into one of a plurality of different thinning patterns. subsampling means for performing subsampling with a thinning pattern selected by a selection signal from , horizontal correlation detection means for detecting the correlation in the horizontal direction of each block data, and vertical correlation detection means for detecting the correlation in the vertical direction of each block data. and a selection signal forming means for forming a selection signal of the thinning pattern from the correlation detection output of the horizontal correlation detection means and the correlation detection output of the vertical correlation detection means, the subsampling means and the selection signal from the selection signal forming means are used as transmission data.