JPH0686076A - Data transmitter - Google Patents

Abstract

PURPOSE:To reduce the entropy per information source symbol by transmitting the information data including a prescribed matrix equivalent to a single block after rearranging the data into a serial information data and inversely scanning this data to rearrange it into a prescribed matrix. CONSTITUTION:The input image data S1 undergoes an orthogonal transformation through a two-dimensional DCT circuit 2 and then is quantized by a quantizing circuit 3. The quantized DCT coefficient data C2 received from the circuit 3 is inputted to the scanning circuits 4A-4C respectively. The circuits 4A-4C turns the data C2 into the run length and input them to a selector circuit 5 as the two-dimensional series data RA1-RA3. At the same time, the EOB symbols EOB1-EOB3 are inputted to a comparator 6. The comparator 6 transmits the scan pattern data STD of the largest 0 run length. Thus the circuit 5 selects one of data RA1-RA3 and inputs it to a variable length coding circuit 8. A multiplexing circuit 7 turns the data RA0 into an entropy code, and the entropy code is multiplexed together with the data STD.

Description

Detailed Description of the Invention

[0001]

[Table of Contents] The present invention will be described in the following order. Field of Industrial Application Conventional Technology Problems to be Solved by the Invention Means for Solving the Problems (FIG. 1) Action (FIG. 1) Embodiments (FIGS. 1 to 5) Effects of the Invention

[0002]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a data transmission device,
For example, it can be applied to a device that transmits information data obtained by orthogonally transforming image data.

[0003]

2. Description of the Related Art Conventionally, as a method of encoding image data with high efficiency and transmitting the image data, for example, two-dimensional DCT
Orthogonal transformation is performed by a method such as (discrete cosine transfer),
Some DCT coefficient data obtained as a result are quantized and entropy coded by a variable length coding method such as Huffman coding and transmitted.

The DCT having a two-dimensional matrix as described above
When transmitting one block of coefficient data, after scanning with a scan pattern called so-called zigzag scan scan and converting it into a one-dimensional sequence, the number of consecutive 0s (so-called 0 run) is counted to make a run length, and a value other than 0 As a two-dimensional sequence combined with an amplitude value (which is called non-zero), a transmission method is used in which the entropy per symbol of the information source is reduced by expanding the information source.

If the last coefficient for one block of the DCT coefficient data is 0, the non-zero appearing at the end of the one-dimensional sequence.
If a special code is assigned from the coefficient of to the last zero of the block, for example as the end of block (EOB) symbol, the EOB symbol takes the longest zero run in most cases. For this reason, the length of other 0 runs in the block is shorter than this, and the entropy of the two-dimensional array having 0 runs and non-zero amplitude values is reduced.

[0006]

However, as described above, D
When one block of CT coefficient data is scanned and transmitted by zigzag scanning, EO that takes the longest 0 run in zigzag scanning depending on the distribution of DTC coefficient data.
In some cases, the B symbol cannot be obtained, and there is a problem that the entropy per information source symbol cannot be sufficiently reduced in practical use.

Further, it has been proposed to provide a cut-off at the time of scanning the DCT coefficient data and switch the scanning direction by the energy integration amount. In this case, the length of 0 run taken by the EOB symbol is predetermined, There is a problem that it is not possible to effectively allocate 0 to the EOB symbol and reduce the entropy per information source symbol.

In practice, even if the energy integration amount is selected, if the number of elements of the information source is larger than that of the other scanning patterns, the entropy per one information source symbol is increased as compared with the other scanning patterns. There was a problem that caused it.

The present invention has been made in consideration of the above points, and when transmitting information data in which one block is a predetermined matrix by a simple method and configuration, the entropy per one information source symbol is totally taken. The present invention is intended to propose a data transmission device that can surely reduce the number.

[0010]

In order to solve such a problem, in the present invention, one block of the information data C2 is transmitted in the data transmission device 1 which transmits the information data C2 in which one block is a predetermined matrix. The serial information data RA1 is scanned by different scanning patterns,
A plurality of scanning means 4A, 4A for rearranging to RA2, RA3
B and 4C, a selecting means 6 for selecting a scanning pattern in which consecutive 0s generated in the serial information data RA1, RA2, RA3 obtained from the plurality of scanning means 4A, 4B, 4C are the longest, and the selecting means. The data sending means 5, 7, 8 for sending the serial information data RA0 selected in 6 together with the scan pattern type ST0 are provided.

Further, in the present invention, in the data transmission device 10 for transmitting the information data C2 in which one block is a predetermined matrix, the serial information data RAX10 transmitted together with the scanning pattern ST10 is reversed based on the scanning pattern ST10. A reverse scanning unit 13 that scans and rearranges in a predetermined matrix is provided.

Further, in the present invention, the selection means 6 is the last E of the serial information data RA1, RA2, RA3.
A scanning pattern in which consecutive 0s generated in OB1, EOB2, and EOB3 are longest is selected. Further, in the present invention, the selecting means 6 selects the scanning pattern in units of a predetermined number of blocks. Furthermore, in the present invention, the information data C2 is made to be coefficient data obtained by orthogonally transforming the image data SI.

[0013]

The information data C2 in which one block is a predetermined matrix is scanned by different scanning patterns and rearranged into serial information data RA1, RA2, RA3,
As a result, a scan pattern having the longest continuous 0s in the serial information data RA1, RA2, RA3 or the longest continuous 0s generated at the end is selected and transmitted together with the scan pattern type ST1. Since the serial information data RAX10 transmitted together with the scan pattern is reversely scanned based on the scan pattern and rearranged in a predetermined matrix, the entropy per one information source symbol can be surely reduced as a whole. .

[0014]

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

In FIG. 1, reference numeral 1 denotes a data coding apparatus for coding and transmitting image data as a data transmission apparatus according to the present invention as a whole, and input image data SI.
Is first orthogonally transformed by a two-dimensional DCT (discrete cosine transfer) circuit 2. Subsequently, the DCT coefficient data C1 obtained by the two-dimensional DCT circuit 2 is quantized by the subsequent quantization circuit 3 with a predetermined number of quantization bits.

Quantized DCT sent from the quantization circuit 3
The coefficient data C2 is input to the first, second and third scanning circuits 4A, 4B and 4C. In the first to third scanning circuits 4A to 4C, the quantized DCT coefficient data C
2, as shown in FIGS. 2, 3 and 4, after scanning with different scan patterns and converting them into a one-dimensional sequence,
A non-zero coefficient value and the number of 0s appearing immediately before that are counted to form a run length.

As a result, the first, second and third scanning circuits 4A, 4B and 4C combine these (0
A two-dimensional sequence of run (R) and non-zero coefficient value (A) is created, and an end-of-block (EOB) symbol is assigned to the 0 run appearing at the end of the block as an identification symbol indicating the end of the block. 2D series data RA
1 (Ri, Ai), RA2 (Ri, Ai), RA3 (Ri, Ai) are input to the selector circuit 5, respectively, and respective EOB symbols EOB1, EOB2, EOB3 are input to the comparator 6.

The comparator 6 compares the 0 run lengths of the input EOB symbols EOB1, EOB2 and EOB3 to determine the longest scan pattern of 0 run length, and the result of this determination is used as the scan pattern data ST0. The signal is sent to the circuit 5 and is sent to the multiplexing circuit 8.

As a result, the selector circuit 5 receives the two-dimensional sequence data RA1 (Ri, Ai) from the first, second and third scan circuits 4A, 4B and 4C according to the input scan pattern data ST0. ), RA2 (Ri, A
i) or RA3 (Ri, Ai) is selected, and this is input to the variable length coding circuit 7 as two-dimensional sequence data RA0.

The variable-length coding circuit 7 entropy-codes the input two-dimensional sequence data RA0 by a method such as Huffman coding, and the resulting coded data RAX0 is multiplexed by the multiplexing circuit 8 with the scan pattern data ST0. Output transmission data DOU of the data encoding device 1
Sent as T.

This transmission data DOUT is input as input transmission data DIN to the data decoding device 10 as shown in FIG.
It is separated into AX10 and scan pattern data ST10, of which encoded data RAX10 is decoded by the variable length decoding circuit 12 and sent to the inverse scan circuit 13 as two-dimensional sequence data RA10.

The inverse scan circuit 13 is a variable length decoding circuit 1.
2 from the two-dimensional series data RA10 input from the scan circuit pattern data ST10 input from the separation circuit 11.
Inverse quantization based on the resulting quantized DC
The T coefficient data C12 is inversely quantized by the inverse quantization circuit 14, the DCT coefficient data C10 is inversely orthogonally transformed by the two-dimensional inverse DCT circuit 15, and the image data SO thus obtained is converted into the image data SO of the data decoding apparatus 10. Sent as output.

In the above configuration, for example, the quantized DCT
When the coefficient data C2 is scanned by zigzag scanning as shown in FIG. 2, the two-dimensional series data RA1 (Ri, Ai) is
(0,12) (0,2) (0,3) (0,1) (1,
1) (3, 4) (0, 8) (0, 1) (8, 6) (0,
7) (0, 2) (11, 1) (EOB) becomes EOB
0 runs to 29. At this time, the entropy per symbol of the information source is 0.610, the mean and entropy of 0 runs are 1.91 and 1.585, and the standard deviation σ of 0 runs is 3.546.

Further, as shown in FIG. 3, the same quantized DCT coefficient data C2 is scanned by a vertical scan, the two-dimensional sequence data RA2 (Ri, Ai) is (0,12) (0,
3) (0, 1) (0, 4) (0, 8) (0, 6) (0,
7) (2,1) (0,2) (1,1) (2,2) (0,
1) (EOB), and 0 run of EOB becomes 47.
At this time, the entropy per symbol of the information source is 0.641,
The mean and entropy of 0 runs are 0.416 and 1.041,0
The standard deviation σ of the run is 0.759.

Further, as shown in FIG. 4, similar quantized DCT coefficient data C2 is (0, 12) when the two-dimensional series data RA3 (Ri, Ai) is scanned by a horizontal scan.
(0,2) (0,1) (12,3) (0,1) (13,
1) (0, 4) (0, 8) (13, 2) (0, 6)
(0,7) (0,1) (EOB), and the 0 run of EOB becomes 14. At this time, the entropy per symbol of the information source is 0.598, the average of 0 runs and the entropy is 3.16.
And 1.041, the standard deviation σ of 0 runs is 5.48.

Accordingly, the quantized DCT coefficient data C2 is shown in FIG.
From the matrix shown in FIG.
Two-dimensional series data RA2 (Ri, Ai) in the case of scanning with the vertical scan shown in FIG. 3 is selected. The average of 0 runs of the sequence thus obtained is shorter than the others, and the maximum length of 0 runs is also small. That is, the standard deviation is also minimized, so that the entropy of the run sequence is reduced, the entropy as a whole is reduced, and the compression rate is improved.

2 to 4, the entropy per symbol of each information source is actually obtained as a reference, but this is a value taken by one block and is only a local value.
Rather, it is not the entropy of each block but the variance (standard deviation) of the data inside / outside each block that is important in determining the overall entropy of the block and the average code length after entropy coding.

For example, referring to only the entropy per source symbol in a block, the case of FIG. 4 is the minimum and a horizontal scan pattern must be chosen, which is the opposite of the 0 run standard. The deviation is the largest. That is, the base of the 0 run that should form the Laplace distribution centered on 0 is extended, and the entropy of the two-dimensional sequence as well as the entire run sequence is increased.

Therefore, in this embodiment, the zero deviation contained in the EOB which can easily control the standard deviation of the zero run to be small.
The scan pattern is selected according to the number of .times. And the two-dimensional sequence data RA0 is entropy coded by the variable length coding circuit 8 to shorten the average code length.

According to the above configuration, the image data SI is set to 2
When transmitting the DCT coefficient data obtained by the three-dimensional DCT conversion, the DCT coefficient data C2 formed by a predetermined matrix is scanned by different scan patterns, and two-dimensional of 0 run (R) and non-zero coefficient value (A). By making an array and assigning an EOB symbol to the 0 run that appears at the end of the block, and selecting and transmitting a scan pattern according to the length of the 0 run included in the EOB symbol, the entire data is transmitted. It is possible to realize the data encoding device 1 and the data decoding device 10 that can surely reduce the entropy per information source symbol.

In the above embodiment, the case where the scanning pattern is controlled in units of 1 block to reduce the entropy of the whole is described, but instead of this,
The optimum scanning pattern may be determined according to the EOB of several blocks, and control may be performed in units of several blocks. By doing so, the overhead can be reduced as a whole.

Further, in the above embodiment, 0 of EOB is set.
The case where the skiyan pattern with the longest run is selected has been described. However, instead of the 0 run of EOB, (0 run, non-zero
Even if the Sukiyan pattern having the smallest 0 run maximum of 0) or the Sukiyan pattern having the smallest 0 run standard deviation or variance is selected, the same as in the above-described embodiment. The effect can be realized.

Further, in the above-mentioned embodiment, the case of transmitting the DCT coefficient data obtained by performing the two-dimensional DCT conversion of the image data has been described. Even if the orthogonal transform method is used, the same effect as that of the above-described embodiment can be realized.

Further, in the above-mentioned embodiment, the case where three types of scan patterns of zigzag scan scan, vertical scan scan, and horizontal scan scan are used as scan scan patterns is described, but the scan scan pattern according to the present invention is not limited to this. Even if the scan pattern is used, the same effect as that of the above-described embodiment can be realized.

Further, in the above-described embodiments, the present invention has been described with respect to the case where the image data is orthogonally transformed and the resulting matrix-shaped coefficient data is scanned and transmitted. However, the present invention is not limited to this. It is suitable for wide application when scanning and transmitting information data in which one block has a predetermined matrix shape.

[0036]

As described above, according to the present invention, information data of which one block is a predetermined matrix is scanned by different scanning patterns and rearranged into serial information data, and as a result, the last of the serial information data is obtained. The longest number of 0s generated in the tail is selected and sent out together with the type of the scanning pattern. The serial information data thus transmitted together with the scanning pattern is reverse-scanned based on the scanning pattern to obtain a predetermined number. By rearranging in a matrix, it is possible to realize a data transmission device that can surely reduce the entropy per information source symbol as a whole.

[Brief description of drawings]

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

FIG. 2 is a schematic diagram used for explaining zigzag skiyan executed by a scan circuit of the data encoding device in FIG.

3 is a schematic diagram for explaining a vertical scan executed by a scan circuit of the data encoding device of FIG. 1. FIG.

4 is a schematic diagram for explaining a horizontal scan executed by a scan circuit of the data encoding device in FIG. 1. FIG.

FIG. 5 is a block diagram showing a data decoding device which is an embodiment of a data transmission device according to the present invention.

[Explanation of symbols]

1 ... Data encoding device, 2 ... Two-dimensional DCT circuit, 3
... Quantization circuit, 4 ... Scan circuit, 5 ... Selector circuit, 6 ... Comparator, 7 ... Multiplexing circuit, 8 ... Variable length coding circuit, 10 ... Data decoding device, 11 ... ... Separation circuit, 12 ... Variable length decoding circuit, 13 ... Inverse scan circuit, 14 ... Inverse quantization circuit, 15 ... Two-dimensional inverse DCT
circuit.

Claims (5)

[Claims] 1. A data transmission device for transmitting information data, one block of which is a predetermined matrix, wherein a plurality of blocks of the information data are scanned by different scanning patterns and rearranged into serial information data. Scanning means, selecting means for selecting the scanning pattern in which consecutive 0s generated in the serial information data obtained from the plurality of scanning means are longest, and the serial information data selected by the selecting means. A data transmission device comprising: a data transmission means for transmitting together with a type of a scanning pattern. 2. A data transmission device for transmitting information data in which one block corresponds to a predetermined matrix, and the serial information data corresponding to one block transmitted together with the scanning pattern is reverse-scanned based on the scanning pattern. Then, the data transmission apparatus is provided with an inverse scanning means for rearranging the matrix into the predetermined matrix. 3. The data transmission according to claim 1, wherein the selecting means selects the scanning pattern in which a continuous 0 generated at the end of the serial information data is the longest. apparatus. 4. The data transmission device according to claim 1, wherein the selecting means selects the scanning pattern in units of a predetermined number of blocks. 5. The data transmission device according to claim 1, wherein the information data is coefficient data obtained by orthogonally transforming an image signal.