JPH057303A - Encoder - Google Patents

Abstract

PURPOSE:To suppress total code data quantity after encoding less definite quantity at the time of encoding picture data. CONSTITUTION:A coefficient processed by an orthogonal tranaforming part 101 is quantized by a quantizing part 102 in conformity to a quantization table 108, and this coefficient is encoded for every hierarchy by hierarch separation, and the code quantity of every hierarchy is counted by a counter part 104 for the code quantity classified by the hierarchy. The hierarchy to be discarded is determined by a code quantity control part 105 based on this information and set coding quantity, and is outputted to a quantization table control part 106. The quantization table to make a coefficient value after the hierarchy to be discarded by this information zero is generated in a quantization table setting part 107, and the coefficient is encoded again by this quantization table so that the total code quantity is suppressed less than the definite quantity. Thus, since sequential code date in which the code quantity is suppressed less than the definite quantity can obtained without converting hierarchical data into the sequential code data, a hardware scale can be miniaturized, and in addition, high-speed processing can be attained.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention is used for an image data storage device, an image data transmission device, etc. for storing and transmitting image data compressed by encoding. The present invention relates to an encoding device to be suppressed below.

[0002]

2. Description of the Related Art When encoding image data such as a still image and compressing the data, attention is paid to the fact that the data of adjacent images have a high correlation, and for example, the image data is divided into blocks of N × N pixels, A method may be used in which data in a block made up of N × N pixels is subjected to orthogonal transform such as discrete cosine transform (DCT), and the transform coefficient is encoded and compressed.

The transform coefficient obtained by applying the discrete cosine transform to the pixels of the N × N two-dimensional block is
In general, for example, in many cases, the energy is concentrated in the low order shown by diagonal lines in FIG. In particular, the transform coefficients of a block having many low frequency components are concentrated in the upper left corner of the block and have a large numerical value, and in the lower right corner, many of them have a small numerical value or zero.

Therefore, when the above transform coefficients are quantized and coded, a variable length coding method is often used to improve the efficiency of data compression by the following processing. That is, for example, when encoding a transform coefficient obtained by performing orthogonal transform on a block of 8 × 8 pixels, coefficient data is sequentially scanned in the direction shown by the arrow in FIG. The set data is formed by combining the number of zero coefficients and the non-zero coefficient value, and the set data is encoded.

More specifically, for example, as shown in FIG. 14, when only the coefficient indicated by the symbol Z is a non-zero coefficient, the number of zero coefficients before the coefficient Z and the coefficient Z are used as group data. The so-called sequential code data is generated by encoding the zero coefficient after the coefficient Z and adding the EOB (End Of Block) code without performing the encoding.

As described above, in the code, the block is a processing unit, and the processing of this block unit is performed on the entire image.
By the way, in the case of accumulating and transmitting image data, there is a demand that the amount of coded data after encoding be kept below a certain amount. For example, when the capacity of the storage medium is limited, it is not desirable that the number of stored sheets be variable depending on the image to be handled. Similarly, when applied to a still camera or the like, it is necessary to guarantee the number of images recorded on a storage medium (magnetic disk, memory card, etc.). Further, in order to deal with moving images, it is necessary to suppress the code amount to be equal to or less than the amount of data that can be read from the storage medium within a fixed time.

As a method of suppressing the code amount below a certain amount, the coefficients after orthogonal transformation are hierarchically encoded for each frequency band, and the code data exceeding a certain amount is code data of the subsequent layers. There is a method in which the total code amount is suppressed to a certain amount or less by truncating. (Japanese Patent Application 3-3
5883) FIGS. 9 and 10 are diagrams for explaining this conventional method. Of these, FIG. 9 shows an example in which the block is divided into three layers (layer A, layer B, and layer C). At the time of encoding, each layer is encoded. FIG. 10A shows a data structure when the entire image is encoded for each layer. With such a code structure, the code amount of each layer can be known. For example, in the figure, if the data amount satisfying the code amount of a certain amount or less is up to layer B, layer C
The code data of is truncated. Next, the code data of the layer A and the layer B in the same block are packed and converted into the code structure of (b). This is because the decoding process is performed in block units, and therefore code data of the same block exists in a dispersed manner because the efficiency becomes poor.

FIG. 11 is a diagram when decoding (b) of FIG.
It is a state of the coefficient in the block. In this way, the coefficient value corresponding to the truncated layer C is processed as zero.

[0009]

However, in the above-mentioned conventional encoding device, when converting to a sequential code in which the total code amount is suppressed to a fixed amount or less, the code of each layer obtained by hierarchically encoding. After calculating the truncation layer from the amount, it is necessary to reconvert the hierarchically stored code data of the valid layer into a sequential code, and for this reconversion, a code memory that holds the hierarchical code data and ,
It is necessary to provide a memory or the like for holding sequential code data that is suppressed to a certain amount or less, and there is a problem that the scale of hardware tends to increase.

In addition, in the conversion process, the code data of each layer stored in the same block in a distributed manner is once decoded, the code length of each variable-length code word is calculated, and the code data of the same block is calculated. There is also a problem that the conversion processing takes a long time because of packing. In view of the above point, the present invention has an object to provide an encoding device capable of performing sequential encoding processing with a total code amount suppressed to a certain amount or less at high speed, and further reducing the scale of hardware. There is.

[0011]

To achieve the above object, a first invention according to claim 1 is an orthogonal transformation means for orthogonally transforming image data divided into N × N pixels, and a first quantization table. And a first table for quantizing orthogonal transformation coefficients in the block with a first quantization table.
And a code amount counting unit for dividing the coefficient data quantized by the first quantization unit into a plurality of layers and counting the sum of the code amounts from the coefficients of the lower layers in each layer. And a second quantum that includes a code amount excess layer detection unit that detects a layer whose count value exceeds the set code amount, and a layer of a layer higher than the layer number detected by the detection unit and that is zero. It is characterized in that it is provided with a quantization table creating means for creating a quantization table, and an encoding means for encoding and transmitting the coefficient data quantized by the second quantization means.

A single second quantization table is created for image data of a predetermined scale such as one screen. Further, the encoding device, when encoding each block coefficient data quantized by the first quantizing means, sets the number of EOB codes assigned to the number of zero coefficients that are continuous last in the block for each layer. And an EOB code counting means for counting the EOB code counting means, wherein the code amount excess layer detecting means has a sum of a count value of the code amount counting means and an EOB code included in a layer to which the count value belongs and a set code amount. It is possible to adopt a configuration in which the hierarchy exceeding the above is detected.

[0013]

According to the present invention, (1) layer information to be truncated is obtained from the code amount of each layer obtained by the first quantizing means according to the above configuration, and the coefficient after the layer to be truncated is set to zero based on this information. By using the second quantizing means for re-encoding, the total code amount is suppressed below a certain amount. (2) The code amount of each layer obtained by the first quantization means,
The layer information to be truncated is obtained according to the number of EOB codes in each layer, and based on this information, a second quantizing means for making the coefficients after the layer to be truncated to zero is used to re-encode the total amount of code. Keep below.

[0014]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An encoding apparatus according to an embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a block diagram of an encoding device according to an embodiment of the present invention.
In FIG. 1, 100 is a memory for storing image data for one screen, 101 is an orthogonal transformation unit, which performs orthogonal transformation on image data of each block obtained by dividing one screen into blocks of 8 × 8 pixels. This is the circuit to apply. FIG. 5A shows coefficient data obtained as a result of performing orthogonal transformation on image data of any one block.

Reference numeral 102 is a quantizer for quantizing the data obtained by the orthogonal transformation. The quantization unit 102 uses the quantization table shown in FIG. 5B held by the quantization table unit 108 to quantize the orthogonal transform coefficient data and the quantization table control unit 106. FIG. 5 newly created by the quantization table setting unit
Two steps including the second step of quantizing the orthogonal transform coefficient data using the quantization table shown in (e) are executed. Each step is continued until all the blocks for one screen stored in the memory 100 are quantized, and when it is completed, the process moves to the second step or the first step of a new image. The configuration of the quantizer 102 is 1
A memory that holds the coefficients that have been orthogonally transformed for blocks,
It is composed of a memory that holds the quantization table shown in FIGS. 5B and 5E, and an arithmetic unit that reads the contents of both memories and performs the quantization processing. Since the memory holding the data obtained by the quantization processing is not included, the quantized coefficient data is output as it is.

A switching unit 103 switches the output direction of the coefficient data quantized by the quantization unit 102, and the output data of the quantization unit 102 is classified by layers until the quantization unit 102 completes the execution of the first step. It is in the state of outputting to the code amount counter unit 104, and after the completion of the execution of the first step, the second
Until the execution of the steps is completed, the output data of the quantization unit 102 is in a state of being output to the encoding unit 109.

The code amount counter unit for each layer 104 divides the quantized coefficient obtained by the quantization unit 102 in the first step into arbitrary layers, and counts the code amount of each layer. The code amount counted by the counter unit 104 is the code amount when entropy coding is performed by the coding unit 109. FIG. 3 shows an example of division within a block. In the figure, the first area is the lowest hierarchy and the fourth area is the highest hierarchy. In addition, although it is divided into four layers in the figure, the number of divisions can be arbitrarily determined.

The code amount control unit 105 detects a layer to be truncated from the code amount of each layer obtained from the code amount counter unit 104 and the set code amount. The quantization table control unit 106 obtains the information on the layer to be cut off obtained from the code amount control unit 105, and creates a table for modifying and controlling the table held by the quantization table unit 108.

The quantization table setting unit 107 is provided with a quantization table 108 while the quantization unit 102 is executing the first step.
The table held by is directly given to the quantization unit 102,
While the quantization unit 102 is executing the second step, a new quantization table (second quantization table) is created based on the table created by the quantization table control unit 106 and the table held by the quantization table 108. It functions to create and give it to the quantizer 102.

The coding unit 109 entropy-codes the quantized coefficient data obtained by the quantizing unit 102 executing the second step, and sends it. Next, the operation of the above configuration will be described. First, when image data for one screen is stored in the memory 100, 8 ×
Image data is read in units of blocks of 8 pixels and input to the orthogonal transformation unit 101. Then, this orthogonal transformation unit 1
At 01, it is converted to the frequency domain as shown in FIG. The transform coefficient is quantized by the quantizer 102. In this case, the quantization is performed by the table held by the quantization table unit 108 (see FIG. 5B).

The coefficient quantized by the quantizer 102 is sent to the code amount counter 104 for each layer, and the code amount counter 104 for each layer counts the code amount of each layer. The code amount count of each layer is performed on all blocks for one screen. The total code amount counts of the respective layers are detected by the code amount control unit 105, and the total count numbers of the code amounts of all the layers are added together, and the count value and the set code amount are compared. The set code amount is a target value for finally suppressing the code data to a predetermined data amount or less. FIG. 4 shows the states of the total code amount and the set code amount of all layers. In the case of the same drawing, the layers that satisfy the set code amount are up to the second layer. Therefore, the code amount control unit 105 notifies the quantization table control unit 106 that the code data of the third and subsequent layers will be truncated based on the code amounts of all the layers and the set code amounts. The quantization table control unit 106 obtains the information on the layer to be truncated from the code amount control unit 105 and generates a quantization table control pattern. FIG. 5D shows an example of the quantization table control pattern. The quantization table control pattern shown in the figure is for the case of truncating code data of the third and subsequent layers. In this case, it is premised that the accuracy of the coefficients after orthogonal transformation is 4 bits. In this way, the coefficient value after quantization is set to zero by setting the table value of the layer to be truncated to a value obtained by adding 1 to the maximum value that can be expressed with the accuracy of the coefficient after orthogonal transformation.

The quantization table setting unit 107 generates a second quantization table from the quantization table control pattern from the quantization table control unit 106 and the quantization table 108. FIGS. 5E and 5F show the second quantization table and the coefficients quantized by this quantization table. In this way, the coefficients of the third and subsequent layers are processed as zero.

When the quantizing unit 102 completes the execution of the first step, the image data is read again from the memory 100 in block units, the orthogonal transforming unit 101 performs orthogonal transform, and the quantizing unit 102 converts it into orthogonal transform coefficients. On the other hand, the second step is executed. That is, quantization is performed using the second quantization table created in the first step, and the quantized coefficient is sent to the encoding unit 103 and encoded. By the above processing, it is possible to obtain a sequential code having a total code amount equal to or less than a certain amount. In the above embodiment, the code amount counter unit 104 separately counts the code amount of each layer in the block, but the present invention is not limited to this, and the code amount is sequentially accumulated from the lower layers. Can also be carried out. In this case, the code amount control unit 105 has the counter unit 1
The layer to be truncated can be detected only by comparing the code amount integrated value obtained from 04 with the set code amount.

FIG. 2 is a block diagram of an encoding apparatus according to another embodiment of the present invention. The configuration of FIG. 2 is basically the same as the configuration of FIG. 1, and detailed description of the same components will be omitted. In the figure, 200 is a memory for storing image data for one screen, 201 is an orthogonal transformation unit for performing orthogonal transformation on the image data divided into blocks, 202
Is a quantizer for quantizing the coefficient obtained by the orthogonal transform unit 201, 203 is an encoder for encoding the coefficient quantized by the quantizer 202, and 204 is a hierarchy for counting the code amount of each hierarchy. A code amount counter unit 206 is a layer-specific EOB code counter unit that counts the number of EOB codes in each layer.
Reference numeral 205 denotes a code amount counter unit 204 for each layer and EOB for each layer.
A code amount control unit 208 for detecting a layer to be truncated based on the code amount of each layer obtained from the code counter unit 206, the EOB code length of each layer and the set code amount and controlling the code amount, 208
Is a first quantization table, 209 is a code amount control unit 205
A quantization table control unit for controlling the quantization table 208 by obtaining the information of the cut-off layer obtained from the above, and a quantization table setting unit 207 for setting the second quantization table from the quantization table 208 and the quantization table control unit 209. And 210 is an encoding unit 20 that actually generates encoded data.
3 is a selector that counts the code amount of each layer and 3 and switches the process of operating the quantization table.

The second quantization table is the memory 100.
Although the image data for one screen stored in 1 is created as one unit, the present invention is not limited to this, and the second quantization table may be created in units of a predetermined amount of data. . Next, the operation of the second coding device will be described with reference to FIGS. 2, 6, 7, and 8.

FIG. 6 is a diagram in which the inside of a block is divided into four hierarchies, all the non-zero coefficients up to the symbol Z in each block, and all the zero coefficients after the symbol Z. FIG. 7 shows an example of coding such a block for each layer. In the figure, the code data of the first layer is (BLK1
-First layer code) (BLK2-First layer code) (B
LK3-the code of the first layer) ..., and the code data of the second layer is (BLK1-the code of the second layer) (BLK2-
Second layer code) (EOB) (BLK3-second layer code) ... and the third layer code data is (BLK3
-Third layer code) (EOB) (BLK3-third layer code) ... and the fourth layer code data is (BLK
(3rd-level code) (EOB) ... In this way, when the EOB code is generated in the Mth layer in the block, code data does not exist in the M + 1th layer and subsequent layers in the same block. That is, there is no more than one EOB in the block. Here, if it is assumed that the hierarchies in which the code amount is equal to or less than a certain amount are code data up to the third layer
The shaded portion of the code data of the fourth layer is cut off and becomes the total code amount. However, if the layered code data from the first layer to the third layer is converted into sequential code data, as shown in FIG.
A code will be added. That is, the code amount in FIG. 8 is larger than the code amount in the shaded area in FIG. 7 by the EOB code length. In this way, when the total code amount is suppressed to a certain amount or less, if the EOB exists in the layer to be cut off, the code amount increases by that amount.

By the way, in the present embodiment, the code amount is counted for each layer, the number of EOBs generated in each layer is counted, and the layer to be temporarily cut off is determined by the code amount of each layer. EO included in valid hierarchy
The number of Bs is calculated, the total number of EOBs existing in the layer to be truncated is calculated from the total number of EOBs that can exist in the entire image, the code amount of the EOB is added to the total code amount of the effective layer, and the set code is added. If the amount is exceeded, the layer to be cut off is 1
Increase one. After that, similar processing is performed and EOB added
Including the above, the layer having the set code amount or less is calculated.

In FIG. 2, the code amount counter unit 2 for each layer is provided.
04, and the EOB code amount information in each layer of the layer-specific EOB code counter unit 206, the code amount control unit 205 uses the above algorithm to perform EOB.
Include a code, find a layer that suppresses the code amount below the set amount,
The quantization table control unit 209 is provided with the cut-off layer information. In this way, a new quantization table is generated by the quantization table setting unit 207, quantization and coding are performed again, and sequential code data in which the total code amount is suppressed to a fixed amount or less is obtained.

[0029]

As described above, according to the present invention,
The coded data encoded using the first quantizing means has code amount and EOB code amount information for each layer, and a layer exceeding a certain amount is detected by this information, and the layers after the layer are set to zero. Since the second quantization table is generated and the quantization and coding processing is performed using the second quantization means,
Conversion processing from hierarchical code data to sequential code is not necessary, and a memory for hierarchical code data is not necessary.

Therefore, the hardware scale can be reduced and the encoding can be performed at high speed.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of an encoding device according to a first embodiment of the present invention.

FIG. 2 is a block diagram showing a configuration of an encoding device according to a second embodiment of the present invention.

FIG. 3 is a diagram showing an example in which a block is hierarchically divided.

FIG. 4 is a diagram showing a state of code data of each layer.

FIG. 5 is a diagram showing a state of coefficients in a block and a quantization table.

FIG. 6 is a diagram showing an example in which a block is hierarchically divided.

FIG. 7 is a diagram showing a state of code data of each layer and code data of each block.

FIG. 8 is a diagram showing a state in which the code of FIG. 7 is converted into a sequential code.

FIG. 9 is a diagram showing an example in which a block is hierarchically divided.

FIG. 10 is a diagram showing how a hierarchical code is converted into a sequential code.

11 is a diagram showing a state of coefficients in a block when the sequential code in FIG. 10 is decoded.

FIG. 12 is a diagram showing a state of coefficients after orthogonal transformation.

FIG. 13 is a diagram showing a scanning direction when the coefficients in a block are arranged one-dimensionally.

FIG. 14 is a diagram showing a scanning direction when the coefficients in a block are arranged one-dimensionally.

[Explanation of symbols]

101 Orthogonal transformation unit 102 quantizer 103 encoding unit 104 Code amount counter unit for each layer 105 code amount control unit 106 quantization table control unit 107 Quantization table setting section 108 quantization table 109 selector

Claims (3)

[Claims] 1. An orthogonal transformation means for orthogonally transforming image data divided into N × N pixels, a table holding means for holding a first quantization table, and a first quantization for an orthogonal transformation coefficient in a block. A first quantizing means for quantizing in a table and coefficient data quantized by the first quantizing means are divided into a plurality of hierarchies, and the sum of the code amounts from the coefficients of the lower hierarchies in each hierarchy Including a code amount counting means for counting, a code amount excess layer detecting means for detecting a layer whose count value exceeds a set code amount, and a layer detected by the detecting means. It is provided with a quantization table creating means for creating a second quantization table that is set to zero, and an encoding means for encoding and transmitting coefficient data quantized by the second quantization means. And an encoding device. 2. The encoding apparatus according to claim 1, wherein a single single second quantization table is created for image data of a predetermined scale such as one screen. 3. The encoding device according to claim 2, further, when encoding each block coefficient data quantized by the first quantizing means, determines the number of zero coefficients that are continuous last in a block. EOB code counting means for counting the number of EOB codes to be allocated for each layer is provided, and the code amount excess layer detecting means is included in the count value of the code amount counting means and a layer after the layer to which the count value belongs. An encoding device for detecting a layer in which the sum total of the EOB code and an EOB code is greater than a set code amount.