JP3263100B2 - Image data encoding method and apparatus - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for encoding a multi-valued image, and more particularly to a method for dividing a multi-valued image into blocks each having a plurality of pixels, and orthogonally transforming the pixels in the blocks. The present invention relates to a method and an apparatus for encoding image data by an orthogonal transform encoding method of a multivalued image to be converted.

In order to store image data whose information amount is orders of magnitude larger than numerical data, especially data of halftone images and color images, or to transmit the data at high speed and high quality, the gradation of each pixel is required. The values need to be encoded efficiently.

[0003]

2. Description of the Related Art As an efficient compression method for image data,
For example, there is an adaptive discrete cosine transform coding method. Adaptive Discrete Cosine Transform Coding
Transform will be described below.

In the ADCT, an image is divided into blocks of 8 × 8 pixels, and an image signal of each block is converted into a spatial frequency distribution coefficient by a two-dimensional discrete cosine transform (hereinafter, referred to as DCT), and is adapted to visual sense. Quantization is performed with the threshold value obtained, and the obtained quantization coefficient is encoded using a Huffman table that is statistically obtained. The encoding operation will be described in detail according to the basic block diagram of ADCT shown in FIG.

First, an image is divided into blocks each having 8 × 8 pixels shown in FIG.
To enter. The DCT transform unit 12 orthogonally transforms the input image signal by DCT, transforms the image signal into DCT coefficients having a spatial frequency distribution shown in FIG. Specifically, as shown in FIG. 11, the image signal input from the terminal 11 is converted by the one-dimensional DCT
The T-transform is performed, and the transposition unit 21 transposes (transposes) the rows and columns of the coefficients in the block and outputs the result to the one-dimensional DCT transform unit 22. The one-dimensional DCT transform unit 22 performs one-dimensional DCT transform in the same manner as the one-dimensional DCT transform unit 20, and outputs the result to the transpose unit 23. The transposition unit 23 performs the same transposition processing as the transposition unit 21 and outputs the result to the terminal 24.

[0006] This process is performed on all blocks of image data to convert them into DCT coefficients. Linear quantization unit 1
In step 3, the input DCT coefficients are linearly quantized by a quantization matrix 14 composed of the quantization threshold values shown in FIG. The result is shown in FIG.

In the quantized DCT coefficient shown in FIG. 15, a DCT coefficient having a value smaller than the threshold value becomes 0, and a quantized coefficient having only the DC component and a small number of AC components is generated.
The quantized coefficients arranged two-dimensionally are converted into one-dimensional by zigzag scan shown in FIG. The variable length coding unit 15 performs variable length coding on the difference between the DC component at the head of each block and the DC component of the previous block. As for the AC component, the variable length encoding is performed on the value (index) of the effective coefficient whose value is not 0 and the length of the run of the invalid coefficient whose value up to the effective coefficient is 0. Further D
Each component of C and AC is encoded using a code table 16 composed of a Huffman table created based on statistics for each image, and the obtained data is output from a terminal 17.

In the ADCT method, the quantization coefficient is DC
It is determined by the T coefficient and the quantization threshold. FIG.
FIG. 3 shows a detailed configuration diagram of a 0 linear quantization circuit 13. FIG.
First, the quantization threshold value storage unit 112 stores a quantization matrix (VTH) as a quantization threshold value. The DCT coefficient input from the input terminal 110 is DC
It is stored in the T coefficient storage unit 111. DCT coefficient storage unit 1
11 outputs the input DCT coefficients to the divider 113 one by one according to the data read signal (RED) from the timing control unit 115.

The quantization threshold value storage unit 112 outputs a quantization threshold value corresponding to each DCT coefficient stored therein to the divider 113 in accordance with the data read signal (RED) from the timing control unit 115. After the quantization of one DCT coefficient is completed, the timing control unit 115
11, and instructs the quantization threshold storage unit 112 to read the next DCT coefficient and quantization threshold, and performs quantization of the next coefficient.

As described above, the DCT coefficients stored in the DCT coefficient storage unit 111 are read out one by one, divided by the quantization threshold stored in the quantization threshold storage unit 112, and the result is calculated by the quantization of the target pixel. By repeating the process of outputting to the terminal 114 as a conversion coefficient for one screen in block units, 1
The DCT coefficients for the screen are quantized.

[0011]

In the ADCT method, when quantizing the DCT coefficients, all the DCT coefficients are quantized with one kind of quantization threshold. In general, human vision tends to relatively hardly recognize deterioration of image quality in an image in which a gradation change greatly changes even if there is some approximation error. The quantization threshold in FIG. 12 utilizes this visibility characteristic.

[0012] However, in contrast, in an image in which the gradation change gradually changes, the human eye tends to recognize the approximation error and tends to recognize the image quality deterioration. Therefore, FIG.
In the case of one type of quantization threshold for an image having a sharp gradation change shown in FIG. 2, deterioration in an image in which the gradation changes gradually is conspicuous, and in order to prevent this, the quantization threshold is set to If the value is set to a low value, the amount of coding when coding an image having a sharp gradation change increases, and there is a problem that efficient compression cannot be performed.

The present invention has been made in view of such a conventional problem, and provides an image data encoding method and apparatus capable of encoding an image of high quality regardless of a gradation change for each image. The purpose is to do.

[0014]

FIG. 1 is an explanatory view of the principle of the present invention, taking an apparatus configuration as an example. First, the present invention
For each block obtained by dividing the original image into a plurality of blocks each including a plurality of pixels, a two-dimensional DCT transform unit 12 for performing a two-dimensional discrete cosine transform of the tone values of the plurality of pixels in the block; A quantization unit 31 for quantizing the transform coefficients obtained from the two-dimensional DCT unit 12 using a predetermined quantization threshold, and an encoding unit for encoding the quantized coefficients obtained by the quantization unit 31 32.

[0015]

According to the present invention, such an image data encoding apparatus is provided with a two-dimensional DCT transform unit 12, a quantization processing unit 31, and an encoding unit 32 for an original image to be processed. Perform the first encoding using the first
The control unit 100 and the total number R _{A of} effective coefficients having a non-zero value up to a predetermined position of each block obtained in the first quantization by the first control unit 100 And the number of effective coefficients R _{B of the} entire image
Ratio R _A / R _B And a two-dimensional DCT transform unit 12 that compares the detection result of the feature detection unit 34 with a predetermined threshold value and uses a quantization threshold value selected based on the comparison result. Control unit 20 that performs the second encoding by the encoding processing unit 31 and the encoding unit 32
0 is provided.

Here, the encoder 32 outputs the first fixed-length code L1 of a fixed fixed length L indicating the ineffective coefficient when the effective coefficient is obtained, for the number of ineffective coefficients until the effective coefficient is obtained. And a fixed-length encoding unit 32a that outputs one second fixed-length code L2 having the same constant length L indicating an effective coefficient;
A variable-length coding unit 32b that performs variable-length coding of a set of the value of the effective coefficient of each component and the number of zeros up to the effective coefficient when the effective coefficients of the component and the AC component are obtained; 10
The first encoding using 0 uses the fixed-length encoding unit 32a, and the second encoding by the second control unit 200 uses the variable-length encoding unit 32b.

The second control means 200 includes the feature detecting unit 3
The ratio R _A / R _B of the number R _B of the total number R _A and the entire image of the effective coefficients of the effective coefficient of up to a predetermined position of each block detected by 4
Is calculated, and the ratio R _A / R _B is compared with a predetermined threshold to select a quantization threshold. Further, the second control unit 200 controls the ratio R
_A / if R _B is less than the threshold value, selects the first quantization threshold applied to intense image gradation change, if the ratio R _A / R _B less than the threshold value, moderate image gradation change Is selected.

On the other hand, the image data encoding method according to the present invention comprises the following processing steps. A first step of performing a first two-dimensional discrete cosine transform, quantization, and encoding on an original image to be processed; an effective coefficient having a non-zero value up to a predetermined position of each block quantized in the first step; Second to detect the total number of effective coefficients and the number of effective coefficients of the whole image
Step: comparing the detection result of the second step with a predetermined threshold, and selecting a quantization threshold based on the comparison result.
Step; 2 using the quantization threshold selected in the third step
A fourth step of performing the two-dimensional discrete cosine transform, quantization, and encoding of the original image for the first time; where the encoding of the quantized coefficient in the first step is divided into whether the quantized coefficient is an effective coefficient or not. In the second step, using the fixed-length code obtained in the first step, the total number of effective coefficients up to a predetermined position in each block is calculated. Detect the number of effective coefficients of the whole image.

[0019]

According to the image data encoding method and apparatus of the present invention having such a configuration, the following operations can be obtained. DCT
In the encoding of image data using, when the coefficients after DCT are quantized with a fixed quantization threshold regardless of the type of image, the number of coefficients remaining as effective coefficients (coefficients other than 0) Varies depending on the magnitude and number of changes in the image.

Generally, in an image such as a landscape image having many portions where the gradation value changes drastically, the number of effective coefficients other than the low order in all the effective coefficients is large, and the gradation change is not significant like a portrait. In the case of an image having a moderate, the ratio of low-order effective coefficients to all effective coefficients tends to increase. Also, the evaluation of the image quality perceived by humans is weak for an image such as a landscape image that changes rapidly, the required approximation accuracy is low, and the coarse quantization is acceptable.

On the other hand, in an image such as a portrait which changes slowly, evaluation tends to be severe, and the required approximation accuracy is high. This can be summarized as follows. [Gradation change] [visual approximation error] [adaptive quantization] Intense (landscape image) Difficult to identify Coarsely good Slow (portrait image) Easier to identify Thinner In the present invention, such image tendency is detected. By performing processing and selecting the optimal quantization conditions based on it, quantization suitable for the characteristics of the image can be performed, and it is possible to stably secure a high-quality restored image while suppressing an increase in the code amount. I do.

That is, in the first encoding, the total number of effective coefficients (coefficients having values other than zero) up to a predetermined position of each block after quantization and the number of effective coefficients of the entire image are detected. A quantization threshold value is selected by comparing the detection result with a preset threshold value, and variable-length coding is performed. FIG.
The operation of the principle diagram of the present invention will be described with reference to the flowchart of FIG.

First, in step S1, an initial quantization threshold, for example, a first quantization matrix VTH1 of a first quantization threshold for performing coarse quantization is selected. Further, in step S2, the fixed length coding unit 32a is selected. These steps S1, S2
May be reversed. Next, in step S3, DCT conversion is performed for each block of 8 × 8 pixels to obtain a DCT coefficient, which is encoded.

At this time, when the encoding processing unit 32 detects an effective coefficient having a value other than 0 from among the one-dimensionally converted quantized coefficients by zigzag scan, the effective coefficient value (index) and the effective coefficient The length of the run of the invalid coefficient (non-effective coefficient) whose value up to 0 is detected. At this time, since the fixed-length coding unit 32a is selected, the fixed-length coding unit 32a converts the invalid coefficient into a first fixed-length code L1 having a fixed code length (fixed length) L bits meaning a coefficient by the number of runs. After that, one second fixed-length code L2 having a constant code length of L bits meaning an effective coefficient is output. The process of step S3 is repeated for each block until the end of all blocks is determined in step S4.

When the first encoding process is completed for one screen, the process proceeds to step S5, where the feature detecting unit 34 reaches a predetermined position of each block based on the fixed length code L2 indicating the effective coefficient stored in the buffer. The feature detection for detecting the total number of effective coefficients having a value other than 0 and the number of effective coefficients of the entire image is performed. Here, the output code data amount M stored in the buffer has a size of M = L × {(number of invalid coefficients) + (number of valid coefficients)} = L × (number of pixels).

The code data thus obtained in the buffer has a code length of L × 64 bits, that is, a code length of L × 8.
For each byte, the identification code L2 of the effective coefficient and the identification code L1 of the invalid coefficient are arranged in the order in which each block is zigzag scanned. The feature detection, and detects the number R _B of the total number R _A and the entire image of the effective coefficient of effective coefficients included in the low-order area from the top scan order of the zigzag scan to N-th, the ratio R _A / R _B Ask for.

[0027] More specifically, to detect the number r _b having r _a and the block overall effective coefficient of effective coefficients up to N-th for each block, which obtain each sum asking per all blocks.
That is, determining the ratio R _A / R _B according to the following equation.

[0028]

(Equation 1)

Based on such a feature detection result of the image, in step S6, one of two or more quantization thresholds prepared is selected, and the quantization threshold value for instructing the quantization processing unit 31 to set the quantization threshold is selected. Is determined by the ratio _RA / R based on feature detection.
If _B is less than the threshold value, selects the first quantization threshold applied to intense image gradation change, if the ratio R _A / R _B less than the threshold value, is applied to loose image gradation change A second quantization threshold value is selected.

After the selection of the quantization threshold value, in step S7, the coding processing unit 32 is instructed to select the variable length coding unit 32a, and in steps S8 and S9, the second DCT transformation, quantization and coding are performed. . At this time, the quantized coefficients arranged two-dimensionally are converted into one-dimensional by zigzag scan,
The DC component and the AC component are each subjected to variable-length encoding by the variable-length encoding unit 32b, and output as formal code data via a buffer.

[0031]

FIG. 3 is a block diagram showing an embodiment of the present invention. In FIG. 3, reference numeral 12 denotes a two-dimensional DCT transform unit,
31 is a quantization processing unit, 32 is an encoding processing unit, 33 is a code buffer, 34 is a feature extraction unit, 35 is a first control unit 100 that performs the first encoding process, and 35 is a second control unit that performs the second encoding process. 2 is a control unit having a function as the control unit 200.

Here, the quantization processing section 31 includes a first quantization threshold value storage section 42 storing a first quantization matrix VTH1 for roughly quantizing DCT coefficients, and a second quantization matrix VTH2 for finely quantizing DCT coefficients. Is stored in a second quantization threshold value storage unit 43 storing the data, a selection unit 41 for selecting two types of quantization threshold values, and a linear quantization unit 40 for linearly quantizing the DCT coefficients using the selected quantization threshold value. .

The encoding unit 32 includes a variable-length code table 53 storing variable-length codes, a characteristic detection code table 52 storing fixed-length codes used for image characteristic detection, and a control unit 35. The selection unit 51 selects two sets of code tables 52 and 53 according to the instruction. Further, the image feature detection unit 3
4 is a fixed-length code L2 indicating an effective coefficient having a value other than 0
Number nonzero bit detector 61 which detects the entire register 62 for counting the number R _B of the effective coefficients of the entire image, the low-order register for counting the number R _A low-order effective coefficients up to a predetermined position 63, and constitutes at the comparison operation unit 64 for comparing the calculation with a predetermined threshold value of the ratio R _a / R _B using the value stored in both registers 63 and 64.

Next, the operation of FIG. 3 will be described with reference to the flowchart of FIG. The control unit 35 starts the first encoding process in order to determine the properties of the image. First, as an encoding condition, in step S1, an initial quantization threshold VT is set.
H is set. For example, the selection unit 41 selects the first quantization matrix VTH1 used for coarse quantization in the first quantization threshold value storage unit 42 and sets it in the linear quantization unit 40.

In step S2, the selection unit 51 selects a feature detection code table (fixed length coding table) 52 in order to perform fixed length coding. Further, in step S3, the code buffer 33 is initialized and all the values are set to 0. The operations in steps S1 to S3 may be performed in any order. Next, step S4
To perform the first encoding. That is, the image is 8 × 8
Divided into pixel blocks and two-dimensional DCT for each block
The DCT is performed by the transform unit 12, and then the linear quantization unit 40
Then, the DCT coefficient is quantized using the first quantization matrix initially set as described above, and the variable-length coding unit 50 performs fixed-length coding using the feature detection code table 52.

At this time, the quantized coefficients arranged two-dimensionally as shown in FIG. 15 are converted to one-dimensional by zigzag scan shown in FIG. Variable length coding unit 50
Indicates the length of a run of an effective coefficient (index) whose value is not 0 and a run length of an invalid coefficient (emergency coefficient) whose value up to the effective coefficient is 0 among the one-dimensionally converted quantized coefficients. Then, a fixed-length code is created with reference to the feature detection code table 52. The feature detection code table 52 has, for example, the contents shown in FIG.

In the first encoding, the variable-length encoding unit 50 functioning as a fixed-length encoding unit, when an effective coefficient whose value is not 0 is obtained, executes a run value indicating the number of invalid coefficients whose value is 0. A constant code length L, for example, L = 1 bit value is 0
After the first fixed-length code L1 is output, one fixed-length code L representing the effective coefficient, and one second fixed-length code L2 in which the value of L = 1 bit is 1, are output to the code buffer 33. Store.

Assuming that the quantization coefficient of the first block shown in FIG. 15 is obtained, the code buffer 33 is arranged in a one-dimensional manner by zigzag scanning and then fixed-length coded as shown in FIG. .. 000 "is stored.

FIG. 7 shows the head block of FIG. 6 taken out. If 8 bits are 1 byte, one block is composed of 8 × 8 = 64 pixels. A first fixed length code L1 (= 0) indicating an effective coefficient having a length L = 1 or a second fixed length code L2 (= 1) indicating an invalid coefficient is arranged. If it is determined in step S4 that the encoding process in step S3 has been performed for all coefficients in all blocks, the process proceeds to step S5.

In step S5, the contents are read from the first bit of the code data stored in the code buffer 33 by the feature detection unit 34, and the processing of steps S6 to S10 is executed to read the contents of each block stored in the coding buffer. performs several detection of R _B number of the total number R _a and the entire image of the effective coefficients of the effective coefficient of up to a predetermined position. In this embodiment, as shown in the zigzag scanning order of FIG.
An example will be described in which the tenth (10th) coefficient is determined as a low-order coefficient from the next coefficient and the number R _{A of} effective coefficients is detected.

First, it is determined in step S6 whether the first bit of the code buffer 33 referred to in step S5 is a valid coefficient. The first one bit is a DC component, which is usually an effective coefficient having a value other than 0.
The general register 62 is incremented by one. Subsequently, in step S8, it is determined whether or not the coefficient is the second to tenth presentation coefficients shown in FIG. Since the first DC component is not included in the low order, the process proceeds to step S10, where the code buffer 33
Is determined, the next lower bit is referenced in step S11.

In the case of the first block shown in FIGS. 6 and 7, since the second is also the effective coefficient of bit 1, step S6,
The process proceeds to step S8 via S7, and since it is a low-order coefficient, proceeds to step S9 to increment the low-order register 63 by one. The above processing is performed for all blocks in the code buffer 33. As a result, the low-order register 63
Is the number of occurrences r of the code 1 indicating the effective coefficient in the data from the second bit to the tenth bit of the block data.
_a, and at the same time, the overall register 62 stores the number of occurrences r _{b of the} code 1 indicating the effective coefficient in all data of the block.
The measured value R _A obtained by accumulating for the entire image by performing this for all blocks to detect the R _B.

When it is determined in step S10 that the processing for all data in the code buffer 33 has been completed, step S10 is performed.
Proceeding to 12, the comparison operation section 64 calculates the ratio of R _A to R _B , H = R _A / R _B, and compares it with a preset threshold value T to determine the image characteristics.

Based on the image feature detection result in step S12, the process proceeds to step S13 or S14, where one of two or more prepared quantization thresholds is selected, and the quantization processing unit 31 corresponds to the selected quantization threshold. To set the quantization matrix to be performed. More specifically, the capacity of the code buffer 33 is 640 pixels × 40 pixels of the target image size.
In the case of 0 pixels, that is, in the case of 4000 blocks, 320
00 bytes.

The comparison operation unit 64 reads the value R _A values R _B and the lower order register 63 for the entire register 62, calculates both of the ratio H = R _A / R _B, for example, of all effective coefficients A comparison is made with a threshold value T = 1/2 for determining whether or not half is in the low order.
The contents of the comparison judgment between the ratio H and the threshold T are as shown in FIG. That is, when the ratio R _A / R _B is smaller than the threshold T, the effective coefficients are dispersed throughout the block, and are a landscape image or the like in which the gradation changes sharply. Since it is difficult to perform quantization, the quantization may be coarse.

On the other hand, when the ratio R _A / R _B is larger than the threshold T, the effective coefficients are concentrated in the lower order of the block, and
It is a portrait with a gradual change in gradation, and a visual approximation error is easy to identify. Therefore, when H <T, the image has a sharp gradation change, and the process proceeds to step S13 to select the first quantization matrix VTH1 of the first quantization threshold value storage unit 42 for coarse quantization.

If H.gtoreq.T, since the image is a slowly changing image, the flow advances to step S14 to select the second quantization matrix VTH2 in the second quantization threshold value storage unit 43 for fine quantization. When the selection of the quantization matrix is completed in this way, the control unit 35 proceeds to step S15, instructs the encoding processing unit 32 to select the variable length code table 53, and in step S16, performs the second DCT transform. The quantization and encoding using the quantization matrix selected in step S13 or S14 are performed until it is determined in step S17 that the processing of all blocks has been completed.

The DCT transformation and quantization in the second quantization are the same as those in the first quantization in step S3, but the quantization of the quantized quantized coefficients is performed by variable length coding. That is, the two-dimensionally arranged quantized coefficients are converted to one-dimensional by the zigzag scan shown in FIG. 16, and the variable length coding unit 50 first calculates the difference between the leading DC component and the DC component of the previous block. Variable length coding is performed by referring to the variable length code table 53.

Subsequently, when an effective coefficient whose AC component value is not 0 is obtained, selection is made in accordance with the value (index) of the effective coefficient and the length of the run of the invalid coefficient whose value up to the effective coefficient is 0. The unit 51 refers to a corresponding position in the variable length code table 53 selected by the unit 51, and outputs data having a different code length recorded at the reference position to the code buffer 33. The variable-length code table 53 used for coding the DC component and the AC component is a Huffman code table created based on statistics for each image.
Consists of a table. The variable-length code data output to the code buffer 33 is finally output from the terminal 17.

In the above embodiment, the fixed-length encoding of the identification code indicating whether the quantization coefficient is an effective coefficient or an invalid coefficient is performed using the feature detection code table 52. A dedicated fixed-length encoding circuit for outputting fixed-length data corresponding to the valid coefficient and the invalid coefficient without using the circuit 52 may be added to the circuit. Further, in the above-described embodiment, an example in which a code having a 1-bit value of 1 is used as the fixed-length code indicating the effective coefficient, but it is apparent that any code length can be used as long as the code has a fixed length. .

Further, in the above-described embodiment, an example has been described in which one of the preset quantization thresholds is selected as the initial quantization threshold for selection at the time of variable-length encoding. The present invention is not limited to this, and can also be realized by using a method using a quantization threshold dedicated to feature detection, or using a result of performing a predetermined operation, for example, addition, subtraction, division or multiplication of a constant, on any of the quantization thresholds. .

[0052]

As described above, according to the present invention, in the first encoding, the total number of coefficients up to a predetermined position of each block after quantization is a non-zero effective coefficient and the effective coefficient of the entire image. By detecting the number of, and comparing the detection result with a preset threshold to judge the nature of the image and selecting an appropriate quantization threshold and performing variable-length encoding, an image having different characteristics can be obtained. On the other hand, it is possible to realize encoding that can always restore a high-quality image with a small code amount.

[Brief description of the drawings]

FIG. 1 is a diagram illustrating the principle of the present invention.

FIG. 2 is a flowchart showing the operation of the present invention.

FIG. 3 is a configuration diagram of an embodiment of the present invention.

FIG. 4 is a flowchart showing the operation of the embodiment of FIG. 3;

FIG. 5 is an explanatory diagram of a feature detection code table used in FIG. 3;

FIG. 6 is an explanatory diagram of a feature detection code string in a code buffer.

FIG. 7 is an explanatory diagram extracting and showing a code string of a first block in FIG. 6;

FIG. 8 is an explanatory diagram of a scanning order of zigzag scanning showing a low-order area according to the present invention.

FIG. 9 is an explanatory diagram of the judgment contents for the magnitude of the ratio R _A / R _B obtained by the feature detection of the present invention.

FIG. 10 is a block diagram of a conventional ADCT coding circuit;

FIG. 11 is a block diagram of the two-dimensional DCT transform of FIG. 10;

FIG. 12 is an explanatory diagram of an original image signal divided into blocks of 8 × 8 pixels.

13 is an explanatory diagram showing DCT coefficients converted from the original image of FIG.

FIG. 14 is an explanatory diagram of a threshold used for quantization of a DCT coefficient;

15 is an explanatory diagram of a quantized coefficient obtained by quantizing the DCT coefficient of FIG. 13 with the threshold of FIG. 14;

FIG. 16 is an explanatory diagram of a scanning order for converting a quantization coefficient into one dimension.

FIG. 17 is a block diagram of a conventional linear quantization circuit.

[Explanation of symbols]

 11.17: Terminal 12: Two-dimensional DCT transformer 31: Quantization processor 32: Encoding processor 32a: Fixed-length encoder 32b: Variable-length encoder 33: Code buffer 34: Feature detector 35: Control Unit 40: linear quantization unit 41, 51: selection unit 42: first quantization threshold storage unit 43: second quantization threshold storage unit 50: variable length coding unit 52: code table for feature detection (fixed length code table) 53: Variable length code table 61: Non-zero bit detection unit 62: Register for the whole 63: Register for low order 64: Comparison operation unit 100: First control unit 200: Second control unit

──────────────────────────────────────────────────続 き Continuation of the front page (56) References JP-A-3-181229 (JP, A) JP-A-3-16489 (JP, A) JP-A-4-286466 (JP, A) JP-A-4-184 247770 (JP, A) JP-A-5-63985 (JP, A) (58) Fields investigated (Int. Cl. ⁷ , DB name) H04N ^7/ 24-7/68

Claims (7)

(57) [Claims] 1. For each block obtained by dividing an original image into a plurality of blocks each consisting of a plurality of pixels, tone values of the plurality of pixels in the block are subjected to two-dimensional discrete cosine transform. In an image data encoding method for quantizing a transform coefficient obtained by the two-dimensional discrete cosine transform using a predetermined quantization threshold, and further coding the quantized coefficient obtained by the quantization, A first step of performing a first two-dimensional discrete cosine transform, quantization, and encoding on the original image, and an effective coefficient having a nonzero value up to a predetermined position of each block quantized in the first step. The total number R _A The ratio R _A / R _B of the number R _B of the entire image of the effective coefficient A third step of comparing the detection result of the second step with a predetermined threshold, and selecting a quantization threshold based on the comparison result; and a quantum step selected in the third step. A fourth step of performing a second two-dimensional discrete cosine transform, quantization, and encoding of the original image using the quantization threshold value. 2. The image data encoding method according to claim 1, wherein the one-time encoding of the quantized coefficient in the first step is performed with different values depending on whether the quantized coefficient is an effective coefficient or not. The fixed-length coding for outputting the fixed-length code is performed. In the second step, the total number of effective coefficients up to a predetermined position of each block and the entire image are calculated using the fixed-length code obtained in the first step. An image data encoding method characterized by detecting the number of effective coefficients. 3. The image data encoding method according to claim 2, wherein the first encoding of the quantized coefficient in the first step includes:
The first fixed length code L1 having a constant length L indicating the ineffective coefficient is output for the number of ineffective coefficients until the effective coefficient is obtained, and the second fixed length code L having the same constant length L indicating the effective coefficient is output.
2 in the second step. In the second step, the total number of effective coefficients from the second fixed-length code L2 obtained in the first step to a predetermined position of each block and the number of effective coefficients of the entire image are calculated. An image data encoding method characterized by detecting. 4. In the image data coding method according to claim 1, wherein the selected quantization threshold in the third process, before SL is obtained a ratio R _A / R _B in the second step is less than the threshold value In this case, select a first quantization threshold applied to an image having a sharp gradation change,
The ratio is when R _A / R _B is not less than the threshold value, the image data encoding method characterized by selecting a second quantization threshold applied to loose image gradation change. 5. A two-dimensional discrete cosine transform for each block obtained by dividing an original image into a plurality of blocks each consisting of a plurality of images, the tone values of the plurality of pixels in the block being two-dimensional discrete cosine transformed. A DCT transformer 12 and the two-dimensional DCT
A quantization unit 31 for quantizing the transform coefficient obtained from the transform unit 12 using a predetermined quantization threshold, and an encoding unit 32 for encoding the quantized coefficient obtained by the quantization unit 31. In the image data encoding apparatus provided with the above, the two-dimensional DCT transform unit 1
2, 1 using the quantization unit 31 and the encoding unit 32
A first controller 100 for performing the first encoding, and a total number R _{A of} effective coefficients having a non-zero value up to a predetermined position of each block obtained in the first quantization by the first controller 100 The ratio R _A between the number R _B of the entire image of the effective coefficient
/ R _B And a feature detection unit 34 that detects the detection result of the two-dimensional DCT transform unit 12 using a quantization threshold value selected based on the comparison result. An image data encoding device, comprising: a second control unit 200 that performs second encoding by a quantization processing unit 31 and an encoding unit 32. 6. An image data encoding apparatus according to claim 5 , wherein said encoding section 32 generates a first fixed length code L1 having a constant length L indicating an ineffective coefficient when an effective coefficient is obtained. The number of non-effective coefficients until the effective coefficient is obtained is output and the second fixed-length code L2 of the same constant length L indicating the effective coefficient is set to 1
And a fixed-length encoding unit 32a that outputs
A variable-length encoding unit 32b that performs variable-length encoding of a set of the value of the effective coefficient of each component and the number of ineffective coefficients up to the effective coefficient when the effective coefficient of the component is obtained; The first encoding by 100 uses the fixed-length encoding unit 32a, and the second encoding by the second control unit 200 uses the variable-length encoding unit 32b. Device 7. The image data encoding device according to claim 5 , wherein the selection of the quantization threshold value in said second control section 200 is performed by setting the ratio R _A / R _B detected by said feature detection section 34 to a threshold value. when it is less than, and selects the first quantization threshold applied to intense image of gradation conversion in the case of more than the ratio R _a / R _B is the threshold value, the applied gentle image gradation change 2. An image data encoding device, wherein a quantization threshold is selected.