JP3617253B2 - Image coding apparatus and method - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image encoding apparatus and method, and more particularly, to enable detection of block noise generated in an encoding process before decoding.
[0002]
[Prior art]
As a method for encoding continuous tone images with high efficiency, an orthogonal transform coding method is widely used. In particular, an algorithm using a discrete cosine transform (hereinafter referred to as DCT) for orthogonal transform is also adopted in JPEG (Joint Photographic Expert Group), which is an international standard for still image coding.
[0003]
FIG. 16 is a block diagram illustrating a configuration of a basic algorithm of JPEG called a JPEG baseline system.
[0004]
In the figure, 100 is an input terminal to which an image is input, 200 is a block dividing unit that divides the input image into pixel blocks that are rectangular areas of 8 × 8 pixels, 300 is subjected to DCT for each pixel block, and conversion coefficients DCT transform unit 400 outputs a quantized coefficient by linearly quantizing the transform coefficient using a quantization step width set for each frequency, and 500 outputs a quantized coefficient. Reference numeral 600 denotes a variable length coding unit that performs variable length coding and outputs code data. Reference numeral 600 denotes an output terminal that outputs code data.
[0005]
FIG. 17 is a block diagram illustrating the configuration of the variable length coding unit 500. In FIG. 17, 101 is an input terminal to which quantization coefficients are input, 501 is rearranged in the order shown in FIG. 10 (zigzag scan), and outputs DC (direct current) components and AC (alternating current) components individually. Scan conversion unit, 502 is a block delay unit that delays the DC component during the processing time of one pixel block, 503 is a subtractor that subtracts the DC component of the block immediately before the DC component and outputs the difference of the DC component, Reference numeral 504 denotes a grouping unit for grouping the difference between the input DC components by absolute values as shown in FIG. 6 and outputting an additional bit indicating the group number SSSS and the numerical value within the group, and 505 is a diagram. A one-dimensional Huffman code that outputs a DC Huffman code by variable-length-coding the DC component group number SSSS using a one-dimensional Huffman code table as shown in FIG. It is a unit.
[0006]
On the other hand, 507 is a zero determination unit that determines whether the input AC component is zero or non-zero and switches the output destination of the AC component. 508 is a run length NNNN that is obtained by multiplying the continuous length of zero coefficients. The run length counter 509 outputs a group of non-zero AC coefficients according to the table of FIG. 7 as in the case of the DC component, and determines the group number SSSS and the AC coefficient in the group. A grouping unit that outputs additional bits to be indicated, and 510 is a two-dimensional Huffman coding that encodes a combination of the run length NNNN and the group number SSSS using the two-dimensional Huffman code table of FIG. 8B and outputs an AC Huffman code Unit 506 multiplexes a DC Huffman code, an AC Huffman code, an additional bit of each of the DC component and the AC component, and outputs code data , 600 is an output terminal of the code data is output.
[0007]
The operation of the JPEG baseline system will be described below with reference to FIG. In FIG. 16, an image input from the input terminal 100 is divided by the block dividing unit 200 into pixel blocks that are rectangular regions of 8 × 8 pixels. FIG. 5A shows an example of a pixel block. The pixel block is subjected to DCT in the DCT conversion unit 300, and conversion coefficients are obtained as an 8 × 8 matrix. The conversion coefficient is information corresponding to the power of the two-dimensional frequency component included in the pixel block.
[0008]
FIG. 4 is a diagram showing the correspondence between each element of the transform coefficient and the frequency. The element at the upper left in the matrix is called a DC component and corresponds to the average value in the pixel block. The other element is called an AC component, and the right component corresponds to a higher frequency in the horizontal direction and the lower component corresponds to a higher frequency in the vertical direction. FIG. 5B is a transformation matrix obtained by applying DCT to the pixel block of FIG.
[0009]
Subsequently, in the quantization unit 400, the transform coefficient is linearly quantized with the quantization characteristic determined for each element.
[0010]
FIG. 13 is a diagram for explaining linear quantization, and shows that the input is quantized with the quantization width Q, and discrete outputs arranged at the interval Q are obtained. The quantization characteristics in JPEG are given, for example, as a matrix shown in FIG. 5C, and quantization is performed by dividing and rounding the conversion coefficient at the corresponding position by these values. FIG. 5D is a quantized coefficient obtained by quantizing the transform coefficient of FIG. 5B with the quantization matrix of FIG.
[0011]
In the variable length coding unit 500, the quantization coefficient in FIG. 5D is variable length coded, and the code data is output to the output terminal 600.
[0012]
Hereinafter, the operation of the variable length coding unit 500 will be described with reference to FIG.
[0013]
The scan conversion unit 501 makes the AC coefficient of the input quantization coefficient one-dimensional in the order shown in FIG. 10 and outputs it. The DC component is output as it is without being processed.
[0014]
The block delay unit 502 stores a DC component during a period in which one pixel block is encoded.
[0015]
The subtracter 503 calculates the difference between the DC component of the current pixel block and the previous DC component stored in the block delay unit 502.
[0016]
As shown in FIG. 6, the grouping unit 504 groups the DC component differences by the magnitude of the absolute value, and outputs an additional bit indicating the group number and the numerical value within the group.
[0017]
The one-dimensional Huffman encoding unit 505 outputs a DC Huffman code corresponding to the group number according to the one-dimensional Huffman code table of FIG. The DC Huffman code is input to the multiplexing unit 506 along with the additional bits.
[0018]
On the other hand, the output destination of the AC component output from the scan conversion unit 501 is switched by the zero determination unit 507 depending on whether it is zero or non-zero.
[0019]
For the AC component determined to be zero, the run length counter 508 counts the continuous length of the zero AC component until a non-zero AC component appears, and outputs the run length.
[0020]
The grouping unit 509 determines a group of non-zero AC components as in the case of the DC component as shown in FIG. 7, and outputs an additional bit indicating the group number and which numerical value the AC component is in the group.
[0021]
As shown in FIG. 9, the two-dimensional Huffman encoding unit 510 performs encoding using a two-dimensional Huffman code table configured by a combination of run lengths and group numbers. Each element in the table is assigned an AC Huffman code as shown in FIG. An AC Huffman code corresponding to the combination of the run length and the group number is output and output to the multiplexing unit 506 along with the additional bits output from the grouping unit 509.
[0022]
FIG. 11A shows a DC component Huffman code and additional bits of the quantization coefficient of FIG. Similarly, FIG. 11B shows an AC component Huffman code and additional bits, and FIG. 11C shows code data obtained from the multiplexing unit 506.
[0023]
With the above procedure, image encoding by the JPEG baseline system is completed.
[0024]
In general, transform coding is lossy coding in which a decoded image does not become the same as the original image. This is because information is lost in the process of quantization, and it is known that noise such as block noise occurs when the compression rate is increased.
[0025]
FIG. 12 is a diagram for explaining block noise. Generation of block noise can be explained as follows. In the transform coding method, since coding is performed for each pixel block, if all AC components are quantized to zero by rough quantization, a block having only an average value is decoded. This is because block boundaries become conspicuous because they are connected with an average value of the surrounding pixel blocks and a step.
[0026]
Block noise is not so noticeable in areas where the pixel amplitude changes drastically, but in areas where the gradation changes slowly, the gradation changes on a rectangular staircase, which is known to cause significant image quality degradation. Yes.
[0027]
Various studies have so far been made on such block noise improvement techniques.
[0028]
As an improvement method on the encoding side, for example, the method disclosed in Japanese Patent Laid-Open No. 4-220083 includes means for detecting a block having only a DC component as block noise, and if detected, a 1-bit flag is added and decoded. In this case, the block is subjected to a smoothing process.
[0029]
Japanese Patent Laid-Open No. 5-328329 switches the quantization characteristic in accordance with the flatness of the pixel block and the AC component at the time of encoding.
[0030]
In Japanese Patent Application Laid-Open No. 7-50758, when a single color document is read, block distortion caused by density fluctuation due to reading error is detected, and a step with a peripheral block is smoothed. Furthermore, when the absolute values of the AC coefficients are all equal to or less than a predetermined value, encoding is performed with all AC coefficients set to zero.
[0031]
As an improvement method on the decoding side, Japanese Patent Laid-Open No. 5-63997 determines the characteristics of the noise removal filter based on the distribution of transform coefficients for each block at the time of decoding. When blocks having only DC components are continuous, noise is removed only when the difference between the DC components of the blocks is equal to or smaller than the quantization step.
[0032]
[Problems to be solved by the invention]
The technique using the block noise detection flag and variable quantization in the prior art can be realized only between specific terminals having these functions, and image transmission using the JPEG baseline algorithm, for example, color There is a problem that it cannot be applied to an application that assumes an unspecified number of receiving terminals such as a facsimile.
[0033]
Further, the conventional technique using the image correction processing at the time of transmission and the adaptive filter after decoding is intended to improve the image quality of the decoded image, and an image faithful to the original image is not necessarily reproduced. Not always.
[0034]
For this reason, depending on the document and application, the image encoded on the transmission side is once decoded and the operator confirms the image quality. If block noise occurs, the image is re-encoded and transmitted with higher image quality parameters. Although it is considered as the most reliable method, human labor is a problem.
[0035]
These problems can be attributed to the fact that the encoding algorithm does not have a function of detecting the occurrence of block noise.
[0036]
Therefore, an object of the present invention is to provide an image encoding apparatus and method having a block noise detection function in order to realize encoding that does not cause block distortion.
[0037]
In addition, in the prior art, for block noise detection,
(1) Detecting a pixel block in which the transform coefficient in the pixel block is quantized to have only a DC component and only the average value is decoded.
(2) The average value of the pixel block is adjacent to the average value of the surrounding pixel blocks with a step.
In many cases, one or both of the above two conditions are used, but there is a problem that a new circuit is required to detect them or the amount of calculation increases.
[0038]
A second object of the present invention is to detect block noise using information obtained in the process of transform coding represented by the JPEG baseline system.
[0039]
[Means for Solving the Problems]
In order to solve the above problems, an image encoding apparatus according to claim 1 of the present invention divides an image to obtain a pixel block which is a rectangular area of N × M (N and M are positive integers) pixels. A dividing means; a transforming means for performing orthogonal transform on the pixel block to obtain transform coefficients; a quantizing means for quantizing the transform coefficients with a predetermined quantization characteristic to obtain quantized coefficients; and a DC of the quantized coefficients The (direct current) component is variable-length-encoded using the difference between the DC component of the immediately preceding pixel block, and the AC (alternating current) component of the quantization coefficient is set to a continuous length of the AC coefficient that is zero and the non-interval that follows this. Two-dimensional variable length encoding means for variable length encoding by a combination of zero AC coefficient magnitudes, The value is +1 or -1. Detecting a first detection signal and outputting a first detection signal; and a continuous length of the AC coefficient The value is NxM-1. And a second detection means for outputting a second detection signal and a determination means for outputting a logical product of the first detection signal and the second detection signal as a detection result for each block, Based on the detection results for each block, When a coarse quantization characteristic is set, it becomes a fine quantization characteristic, and when a fine quantization characteristic is set, it becomes a coarse quantization characteristic. Re-encoding with the quantization characteristic changed is performed.
[0040]
In this configuration, block noise can be detected based on information generated in a conventional encoding process.
[0041]
An image encoding apparatus according to claim 2 of the present invention is the image encoding apparatus according to claim 1, further comprising storage means for storing the second detection signal in the immediately preceding pixel block. Determining means for outputting a logical product of the first detection signal, the second detection signal, and the contents stored in the storage means as a detection result for each block, and based on the detection result for each block Then, the execution of re-encoding with the quantization characteristic changed is determined.
[0044]
Further, the claims of the present invention 3 The image encoding device according to claim 1 is characterized in that in the image encoding device according to claim 1 or 2, the determination means is an AND element.
[0045]
Further, the claims of the present invention 4 The image encoding device according to claim 1 is characterized in that, in the image encoding device according to claim 1 or 2, the orthogonal transform is a discrete cosine transform.
[0046]
Further, the claims of the present invention 5 The image encoding device according to claim 1 is characterized in that, in the image encoding device according to claim 1 or 2, the pixel block is composed of 8 × 8 pixels.
[0047]
Further, the claims of the present invention 6 The image encoding device according to claim 1, wherein the second detection unit has a length of continuous AC coefficients. The value is NxM-1. Instead of detecting the two-dimensional variable length encoding means Output only EOB (End Of Block) code for pixel block with 1 And a second detection signal is output.
[0048]
Further, the claims of the present invention 7 The image encoding device according to claim 1, further comprising a counting unit that counts the detection results for each block in the image encoding device according to claim 1, and when the image encoding operation is completed, Counting result of the counting means of, The total number of pixels of the image or included in the image Find the ratio to the number of all pixel blocks When the counting result is equal to or greater than a predetermined ratio, re-encoding is performed with the quantization characteristic changed.
[0049]
Further, the claims of the present invention 8 An image encoding device according to claim 1 is the image encoding device according to claim 1 or 2, wherein the quantization means includes: D When the quantization threshold of the C component is smaller than a predetermined value, re-encoding with the quantization characteristic changed is not executed.
[0050]
Further, the claims of the present invention 9 An image encoding device according to claim 1 is provided. 8 In the image encoding device described in (1), the predetermined value is determined according to a gradation reproduction characteristic of an output unit that outputs a decoded image.
[0051]
Further, the claims of the present invention 10 The image coding method described in the above section divides an image to obtain a pixel block that is a rectangular area of N × M (N and M are positive integers) pixels, performs orthogonal transformation on the pixel block to obtain a transform coefficient, The transform coefficient is quantized with a predetermined quantization characteristic to obtain a quantized coefficient, and a DC (direct current) component of the quantized coefficient is variable-length-encoded according to a difference between the DC component of the immediately preceding pixel block, and the quantum The AC (alternating current) component of the conversion coefficient is variable-length encoded by a combination of the continuous length of the AC coefficient that becomes zero and the subsequent non-zero AC coefficient, and in particular, The value is +1 or -1. Is detected, the first flag is set, and the AC coefficient has a continuous length. The value is NxM-1. Is detected and the second flag is set, and the first flag and the second flag are set. AND with Based on the detection result for each block, obtain the detection result for each block from When a coarse quantization characteristic is set, it becomes a fine quantization characteristic, and when a fine quantization characteristic is set, it becomes a coarse quantization characteristic. Re-encoding with the quantization characteristic changed is performed.
[0052]
Even in this configuration, it is possible to detect block noise based on information generated in a conventional encoding process.
[0053]
Claims 11 The image encoding method described in claim 10 In the image encoding method according to claim 1, the value of the second flag in the immediately preceding pixel block is set in a third flag, and the first flag, the second flag, and the third flag are set. AND with Then, a detection result for each block is obtained, and based on the detection result for each block, re-encoding with the quantization characteristic changed is executed.
[0054]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the present invention will be described in detail.
[0055]
In the image coding apparatus and method having the block noise detection function of the present invention, block noise is detected based on information obtained in the coding process, and more specifically, information on this coding process. Thus, (1) detection of a pixel block having only an average value and (2) detection of a step with respect to surrounding blocks are performed.
[0056]
Hereinafter, each detection method will be described.
[0057]
A pixel block having only an average value is determined based on whether or not all AC components are zero. In the conventional variable length coding unit 500 shown in FIG. 17, the run length counter 508 counts the run length of the zero AC component, so that the run length output from the run length counter 508 is 63. It can be seen that all AC components are zero.
[0058]
On the other hand, the level difference with the surrounding pixel block can be determined from the group number of the difference of the DC component.
[0059]
In an image region having a gradual gradation change in which block noise is a problem, DC components obtained by applying DCT to a continuous pixel block and quantizing are quantized to the same level or adjacent levels. When quantization is performed on adjacent levels, a difference between levels, that is, a quantization step width becomes a quantization error, and a step is generated at the boundary between decoded pixel blocks.
[0060]
In the case where an edge overlaps the boundary of the original pixel block, the DC component of the pixel block sandwiching the edge is not necessarily quantized to adjacent levels.
[0061]
For this reason, a case where the difference in DC component between consecutive pixel blocks is +1 or −1 can be regarded as a condition for generating block noise. According to FIG. 6, this is nothing but detecting that the group number is 1.
[0062]
In the conventional variable length coding unit 500 shown in FIG. 17, when the group number output from the grouping unit 504 is 1, when the current pixel block is decoded, the boundary with the adjacent pixel block on the left It can be considered that there is a possibility that a step will occur.
[0063]
As described above, according to the present invention, block noise generated in an image area having a gradual gradation change in a document is detected by using these two block noise detection methods in combination.
[0064]
Examples of the present invention will be described below.
[First embodiment]
FIG. 1 shows the first embodiment of the present invention as a whole. In this figure, portions corresponding to those in FIG. In FIG. 1, block noise is detected by a variable length coding unit 500, and the detection result is fed back to the quantization unit 400 to adapt the quantization characteristics.
[0065]
Next, detection of block noise by the variable length coding unit 500 will be described. FIG. 2 shows a variable-length encoding unit 500 of this embodiment. In this figure, portions corresponding to those in FIG.
[0066]
In the figure, reference numeral 511 denotes a group number detection unit that detects a specific group number from the group number SSSS output from the grouping unit 504 and outputs a detection signal. 512 denotes a specific run from the run length NNNN output from the run length counter 508. A run length detection unit 513 that detects the length and outputs a detection signal AND 513 outputs the logical product of the detection signals output from the group number detection unit 511 and the run length detection unit 512 as a block noise detection signal, 601 An output terminal for outputting a block noise detection signal.
[0067]
The group number detection unit 511 detects the case where the DC component difference from the immediately preceding block is +1 or −1, that is, the case where the group number output by the grouping unit 504 is 1, and detects a 1-bit detection signal. Output to AND513.
[0068]
The run length detection unit 512 detects a case where the run length output by the run length counter 508 is 63, and outputs a 1-bit detection signal indicating the detection to the AND 513.
[0069]
The AND 513 outputs the logical product of the detection signals of the group number detection unit 511 and the run length detection unit 512 to the output terminal 601 as block noise detection information.
[0070]
In the first embodiment of the present invention, when block noise is detected from the output terminal 601, the quantization characteristic of the quantization unit 400 is changed and re-encoding is performed. When changing the quantization characteristics, if coarse quantization characteristics are set in advance, the quantization characteristics will be finer, or fine quantization characteristics will be set in advance. Sex If it is set, it is changed so as to obtain a rough quantization characteristic.
[0071]
The decision to execute the re-encoding may be performed by detecting the block noise only once, or the block noise detection signal is counted by a counting unit (not shown), and the counting result is the total number of pixels of the image or the total pixel block. In comparison with the number, it may be determined that the block noise is generated in a predetermined ratio or more.
[0072]
With the above configuration and operation, block noise generated during image coding can be detected by the image coding apparatus according to the first embodiment of the present invention.
[0073]
[Second Embodiment]
Next, a second embodiment of the present invention will be described. In the first embodiment described above, it has not been considered whether the immediately preceding pixel block is a block having only an average value. For this reason, block noise is detected even in the case where a flat pixel block continues to a boundary between an edge region where block noise is relatively inconspicuous and a gentle gradation, for example, a pixel block including an edge.
[0074]
On the other hand, when the detection range is limited to a gradual gradation region, a condition may be added in which the immediately preceding pixel block is a pixel block having only an average value.
[0075]
That is, only when the pixel blocks having only the average value are continuous and a step corresponding to the quantization width of the DC component occurs at the boundary, the block noise is detected.
[0076]
The configuration of the second embodiment of the present invention will be described below with reference to FIG. In addition, about the same structure as the prior art shown in FIG. 11, or the 1st Example of this invention shown in FIG. 2, the same number is attached | subjected and description is abbreviate | omitted.
[0077]
In the figure, reference numeral 514 denotes a storage unit that stores detection information indicating that a specific run length has been detected for the previous pixel block, and reference numeral 515 denotes a detection output from the group number detection unit 511 and the run length detection unit 512. This is an AND that outputs a logical product of the signal and detection information stored in the storage unit 514 as a block noise detection signal.
[0078]
The detection signal output from the run length detection unit 512 is output to the AND 515 and delayed for a period required for encoding one pixel block in the storage unit 514. This detection information is input to the AND 515 together with the detection signals output from the group number detection unit 511 and the run length detection unit 512 in the next block, and the logical product of these is output as a block noise detection signal.
[0079]
With the configuration and operation described above, block noise generated during image encoding can be detected by the image encoding apparatus according to the second embodiment of the present invention.
[0080]
In the JPEG specification, when the run length is 63, the two-dimensional Huffman coding unit 510 outputs an EOB (End of Block) code. For this reason, in the above embodiment, the case where the two-dimensional Huffman encoding unit 510 outputs only the EOB code per pixel block may be detected to determine that all AC components are zero.
[0081]
In the above embodiment, since a pixel block of 8 × 8 pixels is assumed as in the case of JPEG, 63 is detected as a run length when all AC coefficients are zero. However, the present invention is not limited to this value. . For example, when N × M (N and M are positive integers) pixel blocks are used, a run length of (N × M−1) obtained by removing the DC component from the number of transform coefficients (N × M) is detected. That's fine.
[0082]
In addition, the pixel block located at the leftmost end of the image is excluded from the block noise detection process because the immediately preceding block is the block located at the rightmost position in the block line one level above.
[0083]
In addition, since the quantization width of the DC component is relatively small and even if block noise occurs, it may not cause a visual problem. Therefore, if the quantization width of the DC component is smaller than a predetermined value, block noise detection is performed. The function may be stopped or the block noise detection signal may be invalidated. The predetermined value is determined according to reproduction characteristics of an output device such as a display or a printer that outputs a decoded image.
[0084]
Next, the first encoding method of the present invention will be described. FIG. 14 is a flowchart showing the procedure of the encoding method.
[S1] An 8 × 8 pixel block is cut out from the input image.
[S2] DCT is applied to the pixel block to obtain a conversion coefficient.
[S3] The transform coefficient is quantized to obtain a quantized coefficient.
[S4] One-dimensionalize the quantized coefficients in zigzag scan order.
[S5] If the quantization coefficient is a DC component, the process branches to step S6, and if it is an AC component, the process branches to step S11.
[S6] The difference from the DC component of the previous block is calculated.
[S7] The differences are grouped according to FIG. 6 to obtain a group number and additional bits.
[S8] If the group number is 1, the process branches to step S9.
[S9] A flag 1 indicating detection of group number 1 is set.
[S10] The group number is Huffman coded to obtain a DC Huffman code.
[S11] It is determined whether the AC component is zero. If it is zero, the process branches to step S12.
[S12] The run length of zero AC component is counted.
[S13] If the run length is 63, the process branches to step S15.
[S14] AC components are grouped according to FIG. 7 to obtain a group number and additional bits.
[S15] The flag 2 indicating the detection of the run length 63 is set.
[S16] A two-dimensional Huffman code is obtained for the combination of run length and group number.
[S17] The DC Huffman code, the AC Huffman code, and additional bits of both are multiplexed to output code data.
[S18] If flag 1 and flag 2 are set, the process branches to step S19.
[S19] A block noise detection flag is set.
[0085]
Next, the second encoding method of the present invention will be described. FIG. 15 is a flowchart showing the procedure of the encoding method. Steps S1 to S17 are the same as the procedure of the first encoding method, so the same numbers are assigned and description thereof is omitted. Hereinafter, step S18 and subsequent steps will be described.
[S18] If flag 1, flag 2, and flag 3 are set, the process branches to step S19.
[S19] A block noise detection flag is set.
[S20] If flag 2 is set, the process branches to step S21.
[S21] Flag 3 is set.
[0086]
【The invention's effect】
As described above, in the present invention, block noise detection is based on information obtained in the process of the conventional encoding operation such as the difference between DC components and the run length of AC components that become zero. This can be realized by adding a group number detection unit 510, a run detection unit 511, and an AND 513 to the image encoding device. Therefore, it is suitable for mounting on an LSI whose circuit scale is limited or software whose performance is likely to deteriorate due to the addition of noise detection processing.
[0087]
In JPEG, since the code amount varies depending on the contents of the document, various methods for controlling the code amount to a constant value have been devised. For example, the quantization characteristic is changed by multiplying the quantization matrix by a constant value a (a> 0 real number), and a target code amount is converted from the relationship between the generated code amount and a to a convergence operation such as Newton-Raphson method. There is a method to decide by. In this case, since the image quality is not taken into consideration, an image including a large amount of block noise may be decoded although it is the target code amount. By applying the present invention in such a case, it is possible to control the code amount while avoiding block noise.
[Brief description of the drawings]
FIG. 1 is a block diagram showing an overall configuration of a first embodiment of the present invention.
FIG. 2 is a block diagram showing the main part of a first embodiment of the present invention.
FIG. 3 is a block diagram of a second embodiment of the present invention.
FIG. 4 is a diagram illustrating conversion coefficients.
FIG. 5 is a diagram illustrating an example of encoding.
FIG. 6 is a diagram illustrating grouping of DC components.
FIG. 7 is a diagram illustrating grouping of AC components.
FIG. 8 is a diagram illustrating a Huffman code table.
FIG. 9 is a diagram illustrating a two-dimensional Huffman code table for AC components.
FIG. 10 illustrates one-dimensionalization by zigzag scanning.
FIG. 11 is a diagram illustrating an example of variable length coding.
FIG. 12 is an explanatory diagram of block noise.
FIG. 13 is an explanatory diagram of linear quantization.
FIG. 14 is a flowchart for explaining the procedure of the first encoding method of the present invention.
FIG. 15 is a flowchart for explaining the procedure of the second encoding method of the present invention.
FIG. 16 is a schematic configuration diagram of the prior art.
FIG. 17 is a configuration diagram of a variable length coding circuit in the prior art.
[Explanation of symbols]
100, 101 input terminals
200 block division
300 DCT converter
400 Quantizer
500 Variable length coding unit
600 output terminals
601 Block noise detection signal output terminal
501 Scan converter
502 Block delay unit
503 subtractor
504 Grouping Department
505 One-dimensional Huffman encoding unit
506 Multiplexer
507 Zero judgment unit
508 Run length counter
509 Grouping Department
510 Two-dimensional Huffman encoder
511 Group number detector
512 run detector
513 AND
514 storage unit
515 AND

Claims (11)

Dividing means for dividing an image to obtain a pixel block that is a rectangular area of N × M (N and M are positive integers) pixels, transforming means for performing orthogonal transformation on the pixel block to obtain transform coefficients, and the transform coefficients Quantizing means that obtains a quantized coefficient by quantizing with a predetermined quantization characteristic, and variable length coding the DC (direct current) component of the quantized coefficient by the difference between the DC component of the immediately preceding pixel block, A two-dimensional variable length encoding means for performing variable length encoding as a combination of a continuous length of an AC coefficient in which an AC (alternating current) component of a quantized coefficient is zero and a subsequent nonzero AC coefficient size; In an image encoding apparatus,
First detection means for detecting that the value of the difference is +1 or -1 and outputting a first detection signal;
Second detection means for detecting that the continuous length value of the AC coefficient is N × M−1 and outputting a second detection signal;
Determining means for outputting a logical product of the first detection signal and the second detection signal as a detection result for each block;
Based on the detection result for each block, the coarse quantization characteristic is set when the coarse quantization characteristic is set , and the coarse quantization characteristic is set when the fine quantization characteristic is set. and so as, in the image coding apparatus and executes the re-encoded for changing the quantization characteristics. 2. The image encoding apparatus according to claim 1, wherein the second detection signal in the immediately preceding pixel block is stored in the storage unit, the first detection signal, the second detection signal, and the storage unit. Determining means for outputting a logical product of the obtained contents as a detection result for each block, and determining execution of re-encoding with the quantization characteristic changed based on the detection result for each block, An image encoding device. The image coding apparatus according to claim 1, wherein the determination unit is an AND element. The image encoding apparatus according to claim 1, wherein the orthogonal transform is a discrete cosine transform. The image coding apparatus according to claim 1, wherein the pixel block is configured by 8 × 8 pixels. Instead of detecting that the continuous length value of the AC coefficient is N × M−1 in the second detection means, the two-dimensional variable length encoding means performs EOB (End on one pixel block. The image coding apparatus according to claim 1 or 2, wherein the output of only the (Of Block) code is detected and the second detection signal is output. A counting unit that counts the detection result for each block, and when the encoding operation of the image is completed , the total number of pixels of the image or all the pixels included in the image of the counting result of the counting unit; 3. The image encoding device according to claim 1, wherein a ratio with respect to the number of blocks is obtained , and re-encoding is performed by changing the quantization characteristic when the counting result is equal to or greater than a predetermined ratio. 4. . If quantization threshold D C components that put in the quantization means is smaller than a predetermined value, according to claim 1 or 2 characterized in that it does not perform the re-encoding changing the quantization characteristics Image coding apparatus. 9. The image encoding apparatus according to claim 8 , wherein the predetermined value is determined according to a gradation reproduction characteristic of an output unit that outputs a decoded image. The image is divided to obtain a pixel block that is a rectangular area of N × M (N and M are positive integers) pixels, orthogonal transformation is performed on the pixel block to obtain a transform coefficient, and the transform coefficient is determined with a predetermined quantization characteristic. The quantized coefficient is obtained by quantization with a variable length coding of a DC (direct current) component of the quantized coefficient by a difference between the DC component of the immediately preceding pixel block, and an AC (alternating current) component of the quantized coefficient is obtained. In the image encoding method for performing variable length encoding by a combination of the continuous length of the AC coefficient that becomes zero and the size of the subsequent non-zero AC coefficient,
The first flag is set by detecting that the difference value is +1 or −1, and the second value is detected by detecting that the continuous length value of the AC coefficient is N × M−1 . When a flag is set, a detection result for each block is obtained from a logical product of the first flag and the second flag, and a rough quantization characteristic is set based on the detection result for each block Re-encoding is performed to change the quantization characteristic so that the quantization characteristic is fine, and if the fine quantization characteristic is set, the quantization characteristic is coarse. An image encoding method. 11. The image encoding method according to claim 10 , wherein a value of the second flag in the immediately preceding pixel block is set to a third flag, and the first flag, the second flag, and the third flag are set. An image encoding method characterized in that a detection result for each block is obtained from a logical product of and the re-encoding with the quantization characteristic changed is executed based on the detection result for each block.

Images