JPH06233276A - Picture encoder - Google Patents

Abstract

PURPOSE:To dispense with a memory buffer and an adder by executing orthogonal transformation for plural times by the same block in an encoder. CONSTITUTION:The DCT of picture data in a picture memory 42 is executed by a DCT computing element 43 by the scanning of a first time, and the DCT coefficients are quantized by a quantizer 44, and only the effective coefficients are inputted to an entropy encoder 46. A coefficient selector 45 transfers only the quantized DC coefficient C00 by a switching signal from the outside, and inputs the end signal of the transformation coefficients of the block to the entropy encoder 46, and the entropy encoder 46 encodes the transformed coefficients, into entropy and converts them into compressed picture data. By the scanning of a second time, the DCT is executed similarly, and the coefficient selector 45 scans in zigzag fashion the transformation coefficients from a low frequency component including the DC coefficient similarly to the case of sequential encoding, and transfers the effective transformed coefficients, and when all the higher components become ineffective, it inputs the end signal of the transformation coefficients of the block to the entropy encoder 46.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a high-efficiency image coding apparatus using orthogonal transform.

[0002]

2. Description of the Related Art As a redundancy compression / expansion method of a high-efficiency coding / decoding apparatus for television signals, there is an orthogonal transform represented by DCT (discrete cosine transform). Generally, when orthogonal transformation is used for compression / expansion of a television signal, one screen has N
Divide into blocks of pixels × N pixels and perform two-dimensional orthogonal transformation. Here, a signal of one block of N pixels × N pixels is set to [Dij], and a two-dimensional DCT will be briefly described as an example.

If the transformation matrix of the DCT is [Tij], then 2
The three-dimensional DCT is executed by the matrix operation of the following equation to obtain the transform coefficient [Cij].

[Cij] = [Tij] [Dij] [Tij] t ...
.. (1) DCT is a kind of frequency conversion that converts a signal on a frequency axis having no correlation, and when applied to an image signal, effective conversion coefficients concentrate on low frequency components. By utilizing this property, the transform coefficient is transmitted as follows, thereby compressing the redundancy of the image signal.

That is, the conversion coefficient is changed from the low frequency component to C0.
0, C01, C10, C20, C11, C02 ,. ． ． It scans zigzag with CN-1N-1 and transmits up to frequency components at which effective transform coefficients appear, and omits transmission for components above it. For example, if there are valid conversion coefficients up to C20, transmission is performed up to C00, C01, C10, C20, and then E0B (End Of B) indicating the end of the conversion coefficient of the block.
lock) code is transmitted. As a result, the number of transform coefficients to be transmitted is significantly smaller than the number of pixels in the original block, so that the number of transmission bits can be reduced. On the decoding side, the original image data [Dij] is obtained by inverse DCT of the transmitted transform coefficient [Cij].

JPEG (Joint Photographic Experts G
roup) -CD 10918-1 Digital Compression
and Coding of Continuous-tone Still Images Part I:
Requirements and guidelines 3.5 Modes of operation
(March 15, 1991), as a method of DCT-based encoding processing of a still image, in the case of soft copy display such as a display, the following two methods are described from the viewpoint of the display order of the decoded image. Stipulates. The first method is sequential encoding in which all or some of the color components of one image are encoded in one scan from the top to the bottom of the image. On the decoding side, the inverse DCT matrix operation is performed from the transmitted transform coefficient to obtain image data, and the decoded image of the final image quality is sequentially displayed from the top.

The second technique is progressive coding. In this method, all or some of the color components of one image are divided into a plurality of scans and sent. That is, the coefficient of each block is divided into n zones and 1
In the second scan, the coefficients belonging to the first zone are transmitted from the top to the bottom of the image, in the second scan the coefficients belonging to the second zone are transmitted from the top to the bottom of the image,
The same operation is repeated n times. On the decoding side, the resolution is gradually improved after a rough decoded image with low resolution is displayed from top to bottom. This method can convey the outline of the image in a short time when the code amount of the encoded image is large or when it takes time to transmit the code, so that the psychological burden on the decoding side is reduced. There are advantages.

FIG. 1A is an example of a block diagram of a conventional progressive coding apparatus. The image coding device 1 is a coding device of an image coding system using a two-dimensional orthogonal transform, and includes an image memory 2, an orthogonal transformer 3, a quantizer 4, a coefficient memory buffer 5, a coefficient selector 6, It is composed of an entropy encoder 7. 2 (a) is 1
The transmission order of transform coefficients in a block, FIG. 2B is an example of zone division of transform coefficients in progressive coding. In this case, it means that the first zone 21 includes the 0th conversion coefficient, that is, the DC coefficient, the second zone 22 includes the first to 14th conversion coefficients, and the third zone 23 includes the 15th to 63rd coefficients. . The operation of FIG. 1A when the zone division of FIG. 2B is performed will be described. The image data in the image memory 2 is input to the orthogonal transformer 3 and subjected to orthogonal transformation. The transform coefficient of each block is the quantizer 4
Are quantized by and input to the coefficient memory buffer 5 for all blocks in one screen. In the first scan, the coefficient selector 6 operates as shown in FIG.
The reading of the 0th transform coefficient in (b) is requested. Then, the coefficient memory buffer 5 is set to the designated 0 in FIG.
The th transform coefficient is output to the coefficient selector 6. The coefficient selector 6 inputs the read transform coefficient and the end signal of the transform coefficient of the block to the entropy encoder 7. This is repeated for all blocks for one screen. In the second scan, the coefficient selector 6 requests the coefficient memory buffer 5 to read the first to 14th transform coefficients in order from the lowest frequency component in FIG. Then, the coefficient memory buffer 5 outputs the conversion coefficient of the designated order to the coefficient selector 6. The coefficient selector 6 inputs the read transform coefficient and the end signal of the transform coefficient of the block to the entropy encoder 7. This is repeated for all blocks for one screen. In the third scan, the coefficient selector 6 requests the coefficient memory buffer 5 to read the 15th to 63rd conversion coefficients in order from the lowest frequency component in FIG. 2B. Then, the coefficient memory buffer 5 outputs the conversion coefficient of the designated order to the coefficient selector 6. The coefficient selector 6 inputs the read transform coefficients to the entropy encoder 7 until all further components become invalid. After all the further components are invalidated, the end signal of the transform coefficient of the block is input to the entropy encoder 7. This is repeated for all blocks for one screen. The entropy encoder 7 converts the effective transform coefficient into encoded data to obtain compressed image data. FIG. 3 is a diagram showing an example of a coded bit string in conventional progressive coding. Cn0 and Cn1 in the figure indicate the final effective coefficient of each block. In this case, the transform coefficient in one block is divided into three, and an image for one screen is configured by all transform coefficients obtained by three scans.

FIG. 1B is an example of an image decoding apparatus corresponding to the encoding apparatus of FIG. 1A, and includes an entropy decoder 9, an inverse quantizer 10, an inverse orthogonal transformer 11, and an adder 1.
2. The image memory 13 is provided. The operation of the image decoding device 8 will be described. The input compressed image data is input to the entropy decoder 9, and the transform coefficient and its order are decoded. The decoded data, that is, the effective transform coefficient, is input to the inverse quantizer 10 to be inversely quantized, and further input to the inverse orthogonal transformer 11. The result of the inverse orthogonal transform is stored in the image memory 13 as it is in the first scan, is added by the data in the image memory 13 previously transmitted in the adder 12 in the subsequent scans, and is again added in the image memory 13. Stored as reproduced image data. This reproduced image data is read by an image output circuit (not shown) and the reproduced image is displayed. After a rough reproduced image with a low resolution is displayed from the top to the bottom in the first scan, the resolution is sequentially improved from the second scan to the third scan.

The operation of the coefficient selector 6 will be described with reference to FIG. The coefficient selector 6 is composed of an order counter 71, a start order register 72, an end order register 73, a coincidence detector 74, and a coefficient register 75. The data in the start order register 72 becomes the initial value of the order counter 71. The data of the end order register 73 and the order counter 71 are input to the coincidence detector 74, and when they coincide, the block end signal E
OB is output. In the case of the first scan, “0” is input to the start order register 72 and “0” is input to the end order register 73 from the outside. That is, the order counter 71
Is set to an initial value "0" and requests the quantizer 4 of the previous stage to output the 0th (0th) data. The quantized transform coefficient output from the quantizer 4 is input to the coefficient register 75. At this time, since the values of the order counter 71 and the end order register 73 match, the match detector 7
4 outputs a block end signal EOB. Therefore 0th
Since the EOB is output next, that is, at the stage of coding the DC coefficient, only the DC component is coded on the entire screen. In the case of the second scan, the start order register 72 is set to "1" and the end order register 73 is set to "1" from the outside.
14 "is input. That is, the order counter 71 is set to the initial value" 1 ", requests the quantizer 4 of the previous stage to output the first (first) data, and reads the data every time. The read-out quantized transform coefficients are sequentially input to the coefficient register 75. When the order counter 71 reaches "14", the coincidence detector 74 outputs the block end signal EOB. Therefore, EOB is output at the stage of encoding the 1st to 14th data.
Will be output. In the case of the third scan, “15” is input to the start order register 72 and “63” is input to the end order register 73 from the outside. That is, the order counter 71 is set to the initial value "15", requests the quantizer in the previous stage to output the 15th (15th) data, and is incremented to "63" each time the data is read. It The read quantized transform coefficients are sequentially input to the coefficient register 75. The order counter 71 is "
When it becomes 63 ″, the coincidence detector 74 outputs the block end signal EO.
Output B. Therefore, EOB is output at the stage of encoding the 15th to the last effective coefficient.

[0011]

When it is attempted to realize the progressive coding of the above-mentioned prior art, as compared with the case of the above-mentioned sequential coding, the coding device is provided between the quantizer and the entropy coder. It is necessary to add a coefficient memory buffer for the image size. Further, the decoding device requires an adder for accumulating and adding the image transmitted so far and the image data newly transmitted.

It is an object of the present invention to eliminate the memory buffer and the adder and to reduce the scale of the encoding device and the decoding device.

[0013]

In order to achieve the above object, in the present invention, one block of transform coefficients is divided into N zones, and an orthogonal transform is performed in the i-th (1≤i≤N) coding. The coefficients of the first to i-th zones among the coefficients that have been performed are encoded and transmitted for one screen, and are sequentially transmitted from the first encoded data to the N-th encoded data. That is, all the N times of coding are coded from the DC coefficient, and the zone of the coding end, that is, the order of the transform coefficient is increased according to the number of times of coding.

[0014]

On the decoding side, since it is sufficient to decode N kinds of data as independent images of one screen each,
The adder required in the decoding device for progressive coding is not required. In the encoding device, the coefficient memory buffer for the image size becomes unnecessary by performing the orthogonal transformation N times for the same block.

[0015]

EXAMPLE An example of the present invention will be described below with reference to FIG. The image coding device 41 shown in FIG.
An image coding apparatus using a two-dimensional DCT, which includes an image memory 42, a DCT calculator 43, and a quantizer 4.
4, a coefficient selector 45, and an entropy encoder 46. First, a case where this image coding apparatus operates as a sequential coding apparatus will be described. The image data in the image memory 42 is input to the DCT calculator 43, and DCT is performed. The low frequency component of the DCT coefficient is zigzag scanned in the order of the numbers in FIG. 2A, quantized by the quantizer 44, and input to the entropy encoder 46. At this time, the coefficient selector 45 transfers the quantized transform coefficient to the entropy encoder 46, and when all the components above it are invalid, the end signal EOB of the transform coefficient of the block is sent to the entropy encoder 46. Output. The entropy encoder 46 converts the effective transform coefficient into encoded data and outputs it. The image encoding device 47 shown in FIG. 4B corresponds to the image encoding device 41 shown in FIG.
Is a decoding device corresponding to the entropy decoder 4
8, an inverse quantizer 49, an inverse DCT calculator 50, and an image memory 51. The transmitted compressed image data is input to the entropy decoder 48, and the DCT coefficient and its order are decoded. The decoded data, that is, the effective DCT coefficient, is input to the inverse quantizer 49 and inversely quantized, and the inverse quantized data is input to the inverse DCT calculator 50. The result of the inverse DCT is stored in the image memory 51 as reproduced image data.

Next, a case where this image coding apparatus operates as a progressive coding apparatus will be described. Here, as an example, the case where the zone is divided into two, and the first zone has only the DC coefficient and the second zone has all the AC coefficients will be described. In the first scan, the image data in the image memory 42 is input to the DCT calculator 43, and DCT is performed. The DCT coefficients are quantized by the quantizer 44, and only effective coefficients are input to the entropy encoder 46.
At this time, the coefficient selector 45 receives the switching signal from the outside,
Only the quantized DC coefficient C00 is transferred, and the end signal of the transform coefficient of the block is input to the entropy encoder 46. The entropy encoder 46 entropy-encodes the transform coefficient and transforms it into compressed image data. The data transmitted at the time when the first scan is completed is a rough image with low resolution.

In the second scan, the image data in the image memory 42 is input to the DCT calculator 43 again, and D
CT is performed. The DCT coefficient is quantized by the quantizer 44, and the effective coefficient is input to the entropy encoder 46. At this time, the coefficient selector 45 scans the transform coefficient in a zigzag manner from the low frequency component including the DC coefficient in the order of FIG. 2A to transfer the effective transform coefficient, as in the case of the sequential encoding. When all the above components become invalid, the end signal of the transform coefficient of the block is input to the entropy encoder 46. Entropy encoder 46
Entropy-encodes the transform coefficient and transforms it into compressed image data. The data transmitted at the end of the second scan is the compressed image data with the final image quality. At this time, DCT is performed twice for the same block of each color component. An example of compressed image data obtained by the first and second scans is shown in FIG. Cn0 and Cn in the figure
1. ． ． Indicates the final effective coefficient of each block.
At this time, in the image decoding device 47 of FIG. 4B, the compressed image data of the two scans may be decoded as the data of the independent sequential encoding mode. In the above example, an example in which the zone is divided into two has been described, but the same applies when the zone is divided into three or more N pieces. In this case, in the i-th scan, the coefficients of the first to i-th zones will be encoded. Further, DCT is performed N times for the same block of each color component.

Next, the coefficient selector 45, which is a feature of the present invention, will be described with reference to FIG. The coefficient selector 45 is composed of an order counter 81, an end order changeover switch 82, a coincidence detector 83, and a coefficient register 84. Unlike the order counter 71 of FIG. 7, the initial value of the order counter 81 is always "
In addition, the end order changeover switch 82 is different from the coefficient selector in the conventional sequential coding. The input quantized transform coefficient is input to the coefficient register 84 and at the same time the order is changed. The counter 81 counts.The data in the coefficient register 84 is output to the entropy encoder 46 at the next stage and encoded.
In the case of sequential encoding, the end order switch 82 is switched from the outside, and one of the inputs of the coincidence detector 83 is set to "63". That is, the order counter 81
Coincides with "63", the coincidence detector 83 outputs a block end signal EOB. Entropy encoder 4 at the next stage
6 encodes EOB after encoding up to the final effective coefficient. When the data in the coefficient register 84 is valid, it is output to the entropy encoder 46 at the next stage. That is, when the coincidence detector 83 outputs EOB, if the data in the coefficient register 84 is valid at the same time, it indicates that the conversion coefficient of the 63rd order is valid, and if the data is invalid, the coefficient that was valid immediately before that is valid. Up to was valid.
In the case of the first scan of progressive coding, the end order switch 82 is switched and one of the inputs of the coincidence detector 83 is set to "0". That is, when the order counter 81 matches "0", the match detector 83 outputs the block end signal EOB. Therefore, since the EOB is output at the 0th order, that is, the stage where the DC coefficient is encoded, only the DC component is encoded in the entire one screen. In the case of the final scan of the progressive coding, that is, the second scan in the present embodiment, as in the case of the sequential coding, the end order switch 82 is switched and one of the inputs of the coincidence detector 83 is "63". Is set to "" and the final quality image will be encoded.

By the way, in the above embodiment, the case where the same zone division is performed for all color components has been described. However, since it is data for transmitting the outline of the image except for the final scan, all color components are Need not be transmitted.
That is, it is possible to change the zone division for a specific color component and not transmit the color component except for the final scan, or to transmit a constant such as 0. For example, in the case of an image composed of luminance and color difference components, DCT is performed only on the luminance component for the first scan, entropy coding is performed only on the lower order components, and data of color difference component 0 is entropy encoded for the color difference component. Encode. By doing this, it is possible to reduce the load on the DCT calculator and the quantizer.

In the above embodiment, the coding apparatus of the coding system for independently transmitting the DC coefficient of each block has been described. However, the DC coefficient of the block is transmitted by the difference from the DC coefficient of the previous block. The present invention can be implemented with an encoding device. An embodiment of the encoding device in this case is shown in FIG.
FIG. 6B shows an example of the decoding device in FIG.
The difference from the above embodiment is that in the coding device 60, the DC coefficient of the previous block is recorded in the DC coefficient register 61, and at the time of coding, the difference from this coefficient is coded. In 62, the DC coefficient register 6
The point is that the DC coefficient of the previous block is recorded in 3 and the sum with this coefficient is used as the DC coefficient at the time of decoding. Obviously, the other operations are the same as those in the above embodiment.

In the above embodiment, the coding apparatus using the two-dimensional DCT has been described as an example of orthogonal transformation.
The present invention does not depend on the type of orthogonal transformation, but DC
Obviously, the operation will be the same when an orthogonal transform other than T is used.

[0022]

EFFECTS OF THE INVENTION According to the present invention, (1) an adder which has been conventionally required in a decoding device for progressive coding is not required. By performing the orthogonal transform a plurality of times in the same block in the encoding device, the coefficient memory buffer for the image size becomes unnecessary.

(2) An image display effect equivalent to that of progressive coding can be realized by adding a small number of circuits to the coefficient selector of the coding device by sequential coding.

(3) It is possible to realize an image display effect equivalent to that of progressive coding in a decoding device by sequential coding. In addition, the decoding device does not need to recognize the difference between sequential coding and progressive coding.

As described above, the present invention has a large effect in practical use.

[Brief description of drawings]

FIG. 1 is a configuration diagram of an encoding / decoding device of conventional progressive encoding.

FIG. 2 is a diagram showing an example of zone division of transform coefficients in progressive coding.

FIG. 3 is a diagram showing an example of a coded bit string in conventional progressive coding.

FIG. 4 is a configuration diagram showing an embodiment of the present invention.

FIG. 5 is a diagram showing an example of a coded bit string according to the present invention.

FIG. 6 is a configuration diagram showing another embodiment of the present invention.

FIG. 7 is a block diagram showing an example of a coefficient selector in conventional progressive coding.

FIG. 8 is a configuration diagram showing an embodiment of a coefficient selector according to the present invention.

[Explanation of symbols]

2 --- Image memory, 3 --- Orthogonal transformer, 4 --- Quantizer, 5 ---
Coefficient memory buffer, 6 --- coefficient selector, 7 --- entropy encoder, 9 --- entropy decoder, 10--inverse quantizer, 11--inverse orthogonal transformer, 12--adder , 13--image memory,
42--image memory, 43--DCT calculator, 44--quantizer, 45
--Coefficient selector, 46--entropy encoder, 48--entropy decoder, 49--inverse quantizer, 50--inverse DCT calculator,
51--Image memory, 61,63 --- DC coefficient register, 71,81 ---
Order counter, 72--Start order register, 73--End order register, 74,83 --- Match detector, 75,84 --- Coefficient register,
82--End order switch.

Claims (1)

[Claims] 1. A high-efficiency image encoding apparatus using two-dimensional orthogonal transformation, comprising image signal recording means for temporarily storing an input image signal, and read from the image signal recording means. Orthogonal transform means for performing the orthogonal transform a plurality of times on the same image signal, quantizing means for quantizing the output signal from the orthogonal transform means, and externally designating the output from the quantizing means The coefficient conversion means for outputting effective coefficients up to the selected order and means for entropy coding the output from the coefficient selection means. The orthogonal conversion means is a conversion coefficient for one block of image information. Are divided into N zones, the first transformation is performed on the first zone, and the i-th (1 ≦ i ≦ N) encoding is performed by using the coefficients of the first to i-th zones for one screen. Image characterized by encoding Goka apparatus.