JPH09326935A - Image transmission system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To prevent deterioration in a transmission efficiency by suppressing increase in information denoting a quantization table even if the quantization table is frequently varied. SOLUTION: An original image in a frame memory 11 is converted into 8×8 blocks by a block conversion section 12 and the block is subject to discrete cosine transformation (DCT) by a DCT section 13. The data subject to DCT are quantized and compressed by a quantization section 15 by using a proper quantization table T1 decide, based on the result of DCT at a quantization table decision section 14. Furthermore, an entropy coding section 16 conducts Huffman coding to compress the data. A group processing section 18 groups block lines using the same quantization table decided by the quantization table decision section 14 and a parameter addition means adds information of a quantization table to each of one block line or over grouped and the result is sent in a prescribed format.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image transmission system capable of compression-coding a still image in block units and efficiently transmitting the compression-coded code data.

[0002]

2. Description of the Related Art Conventionally, as a typical image compression method, JPEG (Jo
DCT (Discrete) of int Photographic Experts Group
There is a Cosine Transfer method. In this JPEG system, the input image is 8 × 8 (=
64) Divide into blocks of pixels, perform DCT operation, quantization,
Entropy coding (Huffman coding) is sequentially performed. At the time of decoding, contrary to the time of encoding, entropy decoding, inverse quantization, and inverse DCT calculation are performed to irreversibly restore the original input image.

For example, FIG. 6 shows an example of encoding / decoding according to the JPEG-DCT method. An 8 × 8 pixel block B1 of an original image has a DCT coefficient by being subjected to a DCT operation. Converted to data B2. Although the information amount of the converted data B2 itself is not compressed, it is converted into compressed data in which a predetermined high frequency component is deleted by performing quantization using an appropriate quantization table T. The quantization here means the data B of the DCT coefficient.
Dividing each coefficient in 2 by the corresponding coefficient in the quantization table T. This quantized D
The CT coefficient data B3 is converted into further compressed data by performing Huffman coding conversion, and the Huffman coded data and parameters such as a quantization table and a Huffman conversion table are compressed as data. It is transmitted in a predetermined format. When decoding the transmitted compressed data, the receiving side first determines the Huffman coding table using the parameters of the Huffman coding table, and the compressed data B4 received by the Huffman coding table. Decode the Huffman encoded data in. This decrypted data B5
Is the same quantization table T as the quantization table T on the transmitting side.
The inverse quantization is performed by using, and D is almost the same as the DCT coefficient data B2 before quantization on the transmission side.
It is converted into data B6 having CT coefficients. After that, the data B6 is subjected to the inverse DCT conversion, whereby the 8 × 8 pixel block B7 which is almost the same as the 8 × 8 pixel block B1 of the original image can be restored. The JPEG system is described in detail in ITU-T Recommendation (T.81).

The quantization table used at the time of quantization and dequantization in the lossy compression method including the JPEG method includes:
There is no default, an appropriate quantization table can be used for each application, and an appropriate quantization table can be used by switching for each color component forming an image frame. The contents of this quantization table can be changed by changing each coefficient itself in the quantization table or by changing the value of the scaling factor for weighting the quantization. The amount of conversion can be controlled.

Here, if the coding is performed so that the coding amount is constant regardless of the size of the image data, that is, the complexity of the image, the compression rate becomes high in the complicated image area, that is, the quantum. The deterioration of the image quality becomes conspicuous due to the largeization error. On the other hand, in a simple image area,
Since the compression rate becomes low, that is, the quantization error becomes small, the deterioration of the image quality becomes inconspicuous, but the reduction of the encoding amount may not be sufficient.

In order to solve this problem, for example, in the transform coding method disclosed in Japanese Patent Laid-Open No. 2-70127, the entire image is not quantized by one quantization table, but the characteristics of the image are quantized. A coding method has been proposed in which the quantization table is adaptively changed for each block according to the above, quantization is performed, and the coding amount assigned to each divided block is controlled.

[0007]

However, in the transform coding method disclosed in Japanese Patent Application Laid-Open No. 2-70127, information indicating a quantization table necessary for decoding is transmitted with coded data each time. Must be,
When there are many changes in the quantization table, there is a problem that the information indicating the quantization table to be transmitted increases and the overall transmission efficiency decreases.

Therefore, the present invention eliminates such a problem and suppresses an increase in information indicating the quantization table and prevents a decrease in transmission efficiency even when the quantization table is frequently changed. It is an object of the present invention to provide an image transmission system that enables

[0009]

According to a first aspect of the present invention, image data on an image memory is divided into a plurality of blocks, and each of the plurality of divided image data divided in block units is divided into blocks. A predetermined transform coding is performed, the predetermined transform coded data is quantized using a quantization table according to the image characteristics of the divided image data, and a predetermined entropy code is applied to the quantized data. A plurality of entropy-encoded data that has been encoded, and a series of encoding processes are performed, and the plurality of entropy-encoded data and predetermined parameters used in the series of encoding processes are transmitted as compressed data of the image data. Along with
In the image transmission system for decompressing the image data based on the transmitted compressed data, the one or more identifications having the same quantization table applied to each specific area unit composed of one or more blocks Grouping means for grouping areas, and parameter adding means for adding the parameters of the quantization table commonly used for one or more specific areas grouped by the grouping means for each one or more specific area units And is provided.

In a second aspect based on the first aspect, the parameter added by the parameter adding means is acquired, and the acquired parameter is added to the data in the one or more specific areas. It further comprises an assigning means for assigning a quantization table having parameters, and an inverse grouping means for dividing and rearranging one or more specific areas grouped by the grouping means into the specific areas. And

According to a third aspect of the present invention, in the second aspect, the grouping means is arranged so that, when each of the specific areas grouped by the grouping means includes a specific area that is continuous in the arrangement in the image memory, the continuous area. The subgrouping means subgroups the placement information of each specific area into subgroups within the group by using the placement information of the leading specific area. It is characterized in that it has an inverse sub-grouping means for dividing each specified area based on the arrangement information of the head specific area.

[0012]

BEST MODE FOR CARRYING OUT THE INVENTION Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a diagram showing the configuration of an image transmission system according to an embodiment of the present invention. In FIG. 1, the present image transmission system is roughly divided into a transmitter 10 and a receiver 2.
0 and the transmission line 30. Here, the transmission line 3
Reference numeral 0 is a transmission line that communicatively connects the transmission device 10 and the reception device 20, and may be of any form and may be a network.

The transmitter 10 includes a frame memory 11 for holding an original image, a block conversion unit 12, and a DCT calculation unit 1.
3. Quantization table determination means 14 for determining an appropriate quantization table in the quantization table T1, a quantization unit 15, an entropy coding unit 16 for performing Huffman coding based on the Huffman coding table TT1, an entropy code. It has a buffer memory 17 for temporarily storing the converted data, a grouping unit 18, and a parameter adding unit 19.

On the other hand, the receiving apparatus 20 has a parameter recognition unit 21, an entropy decoding unit 22 for Huffman decoding based on the Huffman coding table TT2, and a quantization table T.
2 includes a dequantization unit 24 that performs dequantization using the quantization table, an inverse DCT operation unit 25, an inverse block conversion unit 26, an inverse grouping unit, and a frame memory 28 that holds a restored image. To be done.

Now, with further reference to FIG. 2, the transmitter 1
The encoding of the original image by 0 will be described.

The original image held in the frame memory 11 is held as a frame image shown in FIG. 2A, for example. This frame image may be black and white, or may be an image represented by various color systems. For example, even if it is composed of each color of R (red), G (green) and B (blue) which is an additive color mixture system, C (cyan), M (magenta) and Y (which are additive color mixture systems). Yellow), K
It may be configured in each color of (black).
Furthermore, the YIQ color system as a standard color system for color television broadcasting by the NTSC system may be adopted.

This frame image is transferred to the block conversion unit 12
Is divided into blocks of 8 × 8 pixels (see FIG. 2).
(See (a) and (b)).

The DCT calculator 13 is a block converter 12
DCT calculation is performed for each block divided by.
An 8 × 8 DCT coefficient can be obtained by performing a DCT operation on each pixel in the divided block. The DCT calculation result by the DCT calculation unit 13 is sequentially output to the quantization table determination unit 14 and also to the quantization unit 15. On the other hand, the block conversion unit 12
Then, the arrangement position on the frame image divided into blocks is output to the quantization table determination unit 14.

The quantization table determination unit 14 creates an appropriate quantization table of the quantization table T1 for each specific area in the frame image, here, for each block line, based on the DCT calculation result of each block in the block line. The determined quantization table is sent out to the quantization unit 15 and the scaling factor indicating the determined quantization table or the table number of the determined quantization table is notified to the grouping unit 18. The block line refers to a series of blocks of one line (vertical pixels = 8 pixels) in the line direction (horizontal direction) that form a frame image. The specific area may be, for example, for each block or each frame, other than the block line.

There are the following two methods for determining the optimum quantization table by the quantization table determining section 14, and either method may be adopted, or another method may be adopted. .

The first method is a method of determining the quantization table based on the maximum gradation difference in 64 pixels in the block, that is, the absolute difference between the maximum gradation and the minimum gradation. That is, if this maximum gradation difference is equal to or greater than a predetermined value, it is determined that the block contains characters or edges, and in this case, using a scaling factor that performs weighting so that the quantization error becomes smaller, The coefficient of the standard quantization table is reduced, and the quantization table with the reduced coefficient is adopted as the optimum quantization table. At this time, the scaling factor may be a weighting coefficient of less than 1. Alternatively, the quantization table having the corresponding coefficient is determined from the plurality of quantization tables already weighted by the scaling factor.

The second method is as shown in FIG.
Each of the 64 DCT coefficients in the block is roughly divided into a low frequency component group coefficient (DC, AC0 to AC30) and a high frequency component group coefficient (AC31 to AC62), and the low frequency component group coefficient and the high frequency component group Based on the ratio with the coefficient, the feature of the pixel of the block is detected, and if the detection result is that the coefficient of the low frequency group component is sufficiently larger than the coefficient of the high frequency group component,
Adopting a quantization table in which the coefficient of the quantization table is increased to reduce the code data amount, and conversely, if the coefficient of the low frequency group component is not sufficiently larger than the coefficient of the high frequency group component,
A quantization table in which the coefficient of the quantization table is reduced is adopted so that the code data amount is not reduced. Alternatively, the quantization table having the corresponding coefficient is determined from the plurality of quantization tables already weighted by the scaling factor.

Here, instead of randomly changing each coefficient of the quantization table, one or a plurality of quantization tables to be defaulted as a system are previously set between the transmission device 10 and the reception device 20. It is good to decide. When a scaling factor is used in a system in which one quantization table to be used as a default is used, the quantization table to be applied is uniquely determined only by transmitting the scaling factor. It is advantageous in terms. On the other hand, in a system in which a plurality of default quantization tables are determined, the table number of the quantization table is assigned, and if only the assigned table number is transmitted, the quantization table to be applied is uniquely determined. Therefore, even in this case, it is advantageous in terms of transmission efficiency. Further, in a system in which a plurality of default quantization tables are determined, a scaling factor can be further used. In this case, both the scaling factor to be applied and the table number should be transmitted.

Although the above two methods have been described with respect to the determination of the quantization table for each block, the quantization table determination unit 14 actually determines the optimal quantization table for each block line. , And sends the determined quantization table to the quantization unit 15. As a result, the optimum quantization table can be assigned to each block line, and the optimum quantization table can be changed for each block line.

The quantizing unit 15 performs DCT conversion in each block using the quantization table determined by the quantization table determining unit 14 for each block line composed of blocks having DCT-transformed data. The data is quantized, and the quantized quantized data is output to the entropy coding unit 16. The quantization by the quantizer 15 is
It is realized by dividing the DCT coefficient in each input DCT-transformed block by the coefficient of the corresponding arrangement position in the determined quantization table. As a result, the quantized data has a predetermined high frequency component removed, and the information source in each block is compressed.

The entropy coding unit 16 establishes coding of the quantized data input based on the Huffman coding table TT1 to generate compressed data in which the redundancy of the data is reduced, and the generated compressed data is generated. Buffer memory 1
The data relating to the compressed data output to the buffer memory 17 is output to the grouping unit 18. It is needless to say that the entropy coding unit 16 may perform other entropy coding, not limited to Huffman coding.

When the entropy coding of one image frame is completed, the grouping unit 18 converts the compressed data for each block line whose minimum unit is the block input from the entropy coding unit 16 into the quantization table determining unit 1.
Grouping is performed on the basis of the scaling factor indicating the quantization table applied to each block line input from 4 or the table number of the quantization table. In this grouping, a block line having the same quantization table applied to the block line is taken out from the buffer memory 17, and at least the information of the quantization table and each block line are taken out together with the data of the taken out one or more block lines. Parameter allocation unit 1
This is achieved by sending to 9.

The parameter adding section 19 is a grouping section 1
Scaling factor or table number of the quantization table applied for each one or more block lines grouped by 8 and, if necessary, arrangement information for each block line is added and output to the transmission line 30 in a predetermined format. To do.

When the arrangement information for each block line is necessary, when the block lines having the same quantization table to be applied are consecutive on the image frame, the arrangement information after the second consecutive one is omitted. This is because, at the time of decoding, the second and subsequent consecutive block lines may be simply arranged at the position next to the preceding block line.

The compressed data in which a predetermined format is formed by the parameter adding unit 19 in this manner is sent from the transmitter 10 to the transmission line 30 and transmitted to the receiver 20.

In the receiving apparatus 20, the parameter added by the parameter recognition unit 21 is acquired from the transmitted compressed data. Then, the parameter recognition unit 21
The scaling factor or table number, which is the parameter of the obtained quantization table, is sent to the quantization table determination unit 23, and is sent to the entropy decoding unit 22 in the order of the received data. That is, data in block units are sequentially sent to the entropy decoding unit 22.

The entropy decoding unit 22 determines the same Huffman encoding table TT2 as at the time of encoding based on the parameters of the Huffman encoding table sent from the parameter recognizing unit 21, and determines the determined Huffman encoding table TT2. First, the block-by-block compressed data that is sequentially input is decoded and output to the inverse quantization unit 24.

The quantization table determination unit 23 determines the same quantization table as in the quantization from the quantization table T2 based on the parameters of the quantization table input from the parameter recognition unit 21, and determines the determined quantization table. The quantization table is sent to the inverse quantization unit 24.

The dequantization unit 24 includes an entropy decoding unit 2
The Huffman-decoded block-unit data input from 2 is dequantized using the quantization table determined by the quantization table determination unit 23. The determined quantization table is changed each time the parameter recognition unit 21 acquires and sends it to the quantization table determination unit 23.

The inverse DCT calculator 25 inverse DCT-transforms the block-unit data sequentially input from the inverse quantizer 24, and outputs it to the inverse block transformer 26.

The inverse block conversion unit 26 converts the data in block units sequentially input from the inverse DCT operation unit 25 into a series of image data in which the blocks are deleted. Here, the series of image data is in the order of being grouped by the grouping unit 18. Therefore, the original image cannot be restored with this series of image data.

Therefore, the degrouping unit 27 performs degrouping to replace the image data having the arrangement before being grouped by the grouping unit 18, and the series of image data subjected to this degrouping is set to the frame memory 28.
By outputting to, a restored image that is almost the same as the original image can be obtained.

Next, grouping in block line units will be described with reference to FIG. FIG. 3 shows one frame, where each block line is 1 in the X direction.
It consists of blocks of rows. The number in the Y direction in the frame is arrangement information (location number) indicating the arrangement position on the frame. Therefore, for example, the block line LG1-1 can indicate the position on the frame by the location number “0”.

Here, the block lines LG1-1 to LG-1
1-5 are quantized by the quantization table T1a, and the block lines LG2-1 and LG2-2 are quantized by the quantization table T1b, the block lines LG1-1 to LG1-5 using the quantization table T1a.
Are set to the group of the group LG1 and the block lines LG2-1 and LG2 using the quantization table T1b.
-2 is set to the group of group LG2. In this case, the location number for each block line (0, 1, ... i-1, i, ... K, k + 1, k + 2, ...)
Is not added to all block lines, and only location numbers “0”, “i−1”, and “k” are added. When block lines using the same quantization table are continuous in this way, only the location number of the first block line is added and used as information to be transmitted.

Further, the block lines are grouped and transmitted, but at the time of decoding, the degrouping unit 27 of the receiving device 20 is arranged at the arrangement position on the frame shown in FIG. 3 according to the transmitted location number. Then, each block line is output to the frame memory 28. Therefore, on the frame memory 28 of the receiving apparatus 20, the block numbers are not filled in order from the smallest location number, that is, from the data of the block line with the location number “0”, and the location numbers of the respective block lines are small according to the transmitted group order. Frame memory from the order 2
Will be buried on the top 8. Therefore, the frame memory 2
In the middle of completely filling the data in 8, the block line may be missing.

Next, the data format transmitted from the transmitter 10 to the receiver 20 will be described with reference to FIG.

In FIG. 4, the data of one image is formed by sequentially nesting each data of frame, scan, and group.

The image is S which is the image head marker.
It is composed of an OI (Start of Image), a table, a frame, and an EOI (End of Image) that is an image end marker. The table has a Huffman table definition (DHT: De).
fine Huffman Table) and quantization table specification (DQT:
Required rules such as Define Quantization Table) are inserted.

The frame S is a frame start marker.
It is composed of an OF (Start of Frame), a frame header, and a scan, and various parameters related to the frame are inserted in the frame header like the table. Here, the scan means scan data for each color component when the frame is composed of R, G, and B color components, and in the case of three color components of R, G, and B, scan is performed. It is composed of three scan data of 1, scan 2, and scan 3.

Scan is S, which is a scan start marker.
It is composed of an OS (Start of Scan), a scan header, and a number of groups corresponding to the number of groups grouped by the grouping unit 18. Here, since the number of groups corresponds to the number of used quantization tables, if the number of used quantization tables is 3, there are three groups. Various parameters used for each scan are inserted in the scan header.

The group S is a group start marker.
OG (Start of Group), group header, MCU (Mi
nimum Coded Unit: It is composed of a header and a MCU. The SOG has a code (APP: reserved of APPlication segments) that is reserved for an application in the JPEG standard.
Use the code uniquely assigned from. In the group header, the information of the quantization table used for the quantization of the corresponding group, that is, its value in the case of a scaling factor, and the table of the quantization table when it is selected from a plurality of quantization tables. The number is inserted. The location number is inserted in the MCU header. In the MCU, coded data corresponding to each block is assigned in the minimum coding unit as described above, but since one block line is composed of a plurality of blocks, two blocks are consecutive.
The information on the locations for the second and subsequent blocks is omitted, and the MCU header is not added. Similarly, when the block lines in the group described above are continuous, the addition of the location number is omitted. Therefore, it can be said that the number of MCU headers is the number of groups of discontinuous block lines in the group.

The data format shown in FIG.
The format is based on the JPEG standard, and the MCU data is immediately held in the scan data. On the other hand, in the format according to the embodiment of the present invention shown in FIG. 4, the scan data is first divided into each group grouped by the grouping unit 18, and the group data, that is, the MCU, is divided into each group. Has been inserted.

Therefore, according to the embodiment of the present invention, it is not necessary to repeatedly transmit the information of the used quantization table, that is, it is not necessary to duplicately transmit the information, so that the total amount of information transmitted is reduced. Will be done. In addition, at the time of decoding, since inverse quantization can be continuously performed without using the used quantization table again, efficient quantization can be performed.

In the embodiment of the invention described above, the degrouping unit 27 is arranged after the deblocking unit 26, but the degrouping unit 27 is arranged after the parameter recognition unit 21. The blocks may be arranged, and the grouping may be decomposed at this stage to rearrange the order of the regular block lines in the frame arrangement, that is, the regular block order. However, a memory for that is required.

[0051]

As described in detail above, in the first aspect of the invention, the grouping means applies one or more of the same quantization tables to each specific area unit composed of one or more blocks. Grouping the specific areas, and the parameter adding means adds the parameters of the quantization table commonly used for the one or more specific areas grouped by the grouping means for each of the one or more specific area units; Since the data of one or more specific areas grouped by the grouping means and the parameter of the quantization table added by the parameter adding means are transmitted as at least paired compressed data, the quantization is performed. Even if the table is changed frequently, the increase in the information indicating the quantization table is suppressed and the deterioration of the transmission efficiency is prevented. It has the advantage that it is possible.

In the second invention, the assigning means further acquires the parameter added by the parameter adding means, and the acquired parameter is added to the data in the one or more specific areas to which the parameter is added. , The inverse grouping means restores the image data by dividing and rearranging one or more specific areas grouped by the grouping means into the respective specific areas, and as a result, the grouping is performed. Since decoding is performed every time, there is an advantage that efficient decoding processing can be performed.

In the third invention, the subgrouping means is
When each specific area that is grouped by the grouping unit includes a specific area that is continuous in the arrangement in the image memory, the layout information of each continuous specific area is used in the group by using the layout information of the leading specific area. The sub-grouping means further comprises a reverse sub-grouping means for dividing each specific area sub-grouped by the sub-grouping means based on the arrangement information of the leading specific area. Therefore, there is an advantage that the transmission efficiency can be further improved.

[Brief description of drawings]

FIG. 1 is a diagram showing a configuration of an image transmission system according to an embodiment of the present invention.

FIG. 2 is a diagram illustrating encoding of an original image by the transmission device 10.

FIG. 3 is a diagram illustrating grouping in block line units.

FIG. 4 is a diagram showing a data format transmitted from the transmission device 10 to the reception device 20.

FIG. 5 is a diagram showing a data format based on the JPEG standard.

FIG. 6 is a diagram showing an example of block encoding / decoding according to the JPEG-DCT method.

[Explanation of symbols]

10 ... Transmitting device 20 ... Receiving device 30 ... Transmission path 11, 28 ... Frame memory 12 ... Block conversion unit
13 ... DCT calculation unit 14, 23 ... Quantization table determination unit 15 ... Quantization unit 16 ... Entropy coding unit 17 ... Buffer memory
18 ... Grouping unit 19 ... Parameter addition unit 21 ... Parameter recognition unit 22 ... Entropy decoding unit 24 ... Inverse quantization unit 25
Inverse DCT operation unit 26 Inverse block conversion unit 27 Inverse grouping unit T1, T2 Quantization table

Claims (3)

[Claims] 1. Image data on an image memory is divided into a plurality of blocks, a predetermined conversion encoding is performed on each of the plurality of divided image data divided in block units, and an image of the divided image data is obtained. A sequence of generating a plurality of entropy coded data by quantizing the predetermined transform-coded data using a quantization table according to characteristics and performing predetermined entropy coding on the quantized data. And transmitting the plurality of entropy encoded data and the predetermined parameters used in the series of encoding processing as compressed data of the image data, and based on the transmitted compressed data, In an image transmission system for restoring image data, the quantization table applied for each specific area unit composed of one or more blocks Grouping means for grouping one or more specific areas that are the same, and parameters of the quantization table that are commonly used for one or more specific areas grouped by the grouping means. An image transmission system, comprising: a parameter adding unit for adding each unit. 2. Allocation means for acquiring the parameter added by the parameter adding means, and for allocating a quantization table having the acquired parameter to the data in the one or more specific areas to which the parameter is added The image transmission system according to claim 1, further comprising: and a degrouping unit that divides and rearranges one or more specific regions grouped by the grouping unit into each of the specific regions. . 3. The grouping means, when each specific area grouped by the grouping means includes a specific area that is continuous in terms of arrangement in an image memory, arranges the arrangement information of each continuous specific area at the top. The subgrouping means further subgroups in the group using the arrangement information of the specific area, and the degrouping means places each specific area subgrouped by the subgrouping means at the top of the specific area. 3. The image transmission system according to claim 2, further comprising an inverse sub-grouping unit that divides the image based on the arrangement information of.