JP3855376B2 - Image encoding device - Google Patents

Description

[0001]
BACKGROUND OF THE INVENTION
The present invention relates to an image encoding device, and particularly to lossless encoding for a multi-valued input image.
[0002]
[Prior art]
Since an image generally has a very large amount of data, it is generally compressed by encoding at the time of storage or transmission. Image data encoding methods are roughly classified into two types, lossless encoding and lossy encoding. In the latter case, depending on the compression rate, visual degradation may occur in the decoded image.
[0003]
In recent years, images created by computers (hereinafter collectively referred to as CG images) are increasing, with images described in PDL (Page Description Language) as a representative. Since these images do not contain a noise component at the time of scanning as compared with a conventional scanned-in image, image quality deterioration due to lossy encoding at the edge or the like is very conspicuous.
[0004]
Such a problem of image quality degradation can be avoided by using lossless encoding that can obtain the same decoded image as the input image.
[0005]
The “image coding apparatus and image decoding apparatus” disclosed in Japanese Patent Application No. 95-00161 (currently unpublished) has been made to realize such reversible coding.
[0006]
FIG. 8 is a block diagram showing a configuration of an image encoding / decoding apparatus proposed in Japanese Patent Application No. 95-00161.
[0007]
In the figure, 10 is an image input unit, 20 is a first prediction unit, 21 is a second prediction unit, 22 is a third prediction unit, 30 is a prediction error calculation unit, 40 is a selection unit, 50 is an encoding unit, 60 Is a code output unit. In the selection unit 40, 401, 402, and 403 are subtraction units, 404, 405, and 406 are zero determination units, 407 is AND, 408 is an identification number determination unit, and 409 is a multiplexing unit. .
[0008]
Subsequently, the operation will be described with reference to FIG.
[0009]
The first prediction unit 20, the second prediction unit 21, and the third prediction unit 22 each predict the pixel value of the target pixel based on the image data 100 by a predetermined method, and send the prediction value data 110 to the selection unit 40. .
[0010]
As a prediction method, for example, as shown in FIG. 7A, pixels a, b, c, and d positioned around the target pixel X to be predicted are read, and the values of these pixels are predicted values. Some data is 110.
[0011]
The prediction error calculation unit 30 predicts the pixel value of the target pixel based on the image data 100 by a predetermined method, subtracts the prediction value from the actual pixel value of the target pixel, and outputs the prediction error data 120 to the selection unit 40. Send it out.
[0012]
The selection unit 40 detects a prediction match / mismatch in the target pixel from the image data 100 and the prediction value data 110.
[0013]
Differences between the prediction value data 110 output from the first prediction unit 20, the second prediction unit 21, and the third prediction unit 22 and the image data 100 are calculated in subtraction units 401, 402, and 403, respectively. Further, the zero determination units 404, 405, and 406 determine whether the difference is zero. If the difference is zero, the determination result is notified to the identification number determination unit 408, and if not, the result is notified to the AND 407. The AND 407 detects a case where all the differences calculated by the subtraction units 401, 402, and 403 are not zero, and notifies the identification number determination unit 408 of the result.
[0014]
The identification number determination unit 408 holds an identification number that uniquely identifies each of the four inputs of the determination result output from the zero determination units 404, 405, and 406 and the output of the AND 407, and the identification number corresponding to the input one Is output to the multiplexing unit 409. Note that the assignment of the identification number may be fixed or dynamically changed according to the change in the appearance frequency.
[0015]
The multiplexing unit 409 is one of the predictors of the first prediction unit 20, the second prediction unit 21, and the third prediction unit 22 when any of the determination results of the zero determination units 404, 405, and 406 is zero. Whether or not hit is correct, an identification number corresponding to this is output. When the output of the AND 407 is detected, that is, when any of the predictors of the first predictor 20, the second predictor 21, and the third predictor 22 is not correct, the corresponding identification number and prediction FIG. 6A in which the error 120 is multiplexed and the prediction situation data 130 is output to the encoding unit 50 is a diagram illustrating the configuration of the prediction situation data 130. This is composed of only the identification number of the prediction unit that has been predicted correctly, or an identification number indicating that neither has been hit and the prediction error data 120.
[0016]
The encoding unit 50 performs variable length encoding by replacing the identification number of the prediction situation data 130 with a corresponding variable length code word.
[0017]
FIG. 7B is a diagram for explaining an example of assignment of variable length codes to identification numbers.
[0018]
In the configuration of the predictor in FIG. 7A, when the pixel values of the surrounding pixels a, b, c, and d are used as predicted values as they are, the prediction accuracy is a>b>c>d> (all predictions). And the identification number determination unit 408 determines the identification numbers in descending order of occurrence probability from 0, 1, 2,. . . Is assigned, the code table of FIG. 7B is obtained. The variable-length codeword is determined by configuring a tree of Huffman codes shown in FIG. 7C for the identification numbers arranged in the order of occurrence probability.
[0019]
FIG. 6C shows code data 140 obtained by variable-length coding the prediction status data 130 shown in FIG. 6B.
[0020]
The code output unit 60 outputs the code data 140 to a transmission path (not shown) or storage means.
[0021]
With the above configuration and operation, the image signal is reversibly encoded by the image encoding device in the prior art.
[0022]
[Problems to be solved by the invention]
A CG (computer graphics) image has a feature that (1) the pixel value used is often biased to a specific value, and (2) the same pixel value is often present in an adjacent pixel. Many.
[0023]
In the related technology described above, the efficiency of the CG image is obtained by encoding an identification number of a predictor that predicts the same pixel value as the target pixel by using such characteristics, and by encoding a prediction error when prediction is lost. Lossless encoding was realized. At this time, in order to further increase the coding efficiency, a variable-length code corresponding to the hit rate is assigned to each predictor.
[0024]
However, the related technology has the following problems.
[0025]
For example, consider a case where codeword lengths i, j, and k (natural numbers where i ≦ j <k) are assigned and three predictors are used. If the predictors of codeword length i and j have the same prediction value and these predictions are correct, a codeword of length i with a short code length is always selected from the viewpoint of compression, No codeword is chosen.
[0026]
On the other hand, when the predictors of the codeword lengths i and j have the same predicted value and the prediction is not correct, the predictor to which the codeword length k is assigned may be correct. Although the code word having the length j is not used, the code word having the length k is used as it is.
[0027]
The present invention has been made in view of the above-described circumstances, and an object thereof is to provide a lossless encoding apparatus and decoding apparatus having a high compression rate for computer-generated images including images described in PDL. It is.
[0028]
[Means for Solving the Problems]
According to claim 1 of the present invention, in order to solve the above-described problem, the image input means for inputting an image and the pixel value of the target pixel as an encoding target in the image input by the image input means are different from each other. A plurality of prediction means for predicting by a prediction method; a prediction error calculation means for calculating an error between a pixel value of a target pixel in an image input by the image input means and a prediction value predicted by a predetermined prediction method; A match determination unit that determines whether or not a pixel value predicted by a plurality of prediction units and a pixel value of a target pixel match, and a prediction by any one of the plurality of prediction units by the match determination unit selection means when the pixel value and the pixel value of the pixel of interest that match, outputs the prediction means is determined that the match, if not determined that the match for selecting and outputting the error In the image coding apparatus comprising: encoding means for encoding the identification information and error output by the selection means; and output means for outputting the code encoded by the encoding means, the plurality of predictions It is determined that at least one prediction result of the means coincides and the predicted pixel values of at least two prediction means determined to be inconsistent are equal, and the encoding means includes identification information corresponding to each prediction means. It is characterized by comprising determination means for determining that the assignment of the binary sequence in the encoding means is to be changed based on the length of the given variable-length binary sequence.
[0029]
Further, among the plurality of prediction means, a variable length binary sequence of length k (k is a natural number) is included in the identification information of the prediction means in which the predicted pixel value matches the pixel value of the target pixel. Among the plurality of prediction means, there are a plurality of pixel values predicted by the prediction means in which the predicted pixel value does not match the pixel value of the target pixel, and these When a binary sequence shorter than the length k is assigned in the encoding means, the identification means of the prediction means is shorter than the length k when a notification of binary sequence assignment change is received from the determination means. The pixel value which predicted the longest binary sequence among the binary sequences and the pixel value of the target pixel are assigned to the identification information of the prediction means. Further, in a manner described later with reference to FIG. 3, the length of the binary sequence assigned to the identification information of the prediction means in which the predicted pixel value and the pixel value of the target pixel match is the binary sequence. A binary sequence shorter than the length k allocated to a part of the identification information of each of the plurality of prediction means having the same predicted pixel value so as to be shorter than the length before the allocation change is used as the predicted pixel. You may assign to the identification information of the other prediction means of several prediction means with an equal value.
[0030]
Further, according to claim 2 of the present invention, prior Symbol encoding means and wherein the configuring the code and variable-length coding the binary sequence of Huffman code as a code word.
[0032]
The coincidence determining unit may match the predicted pixel value with the pixel value of the target pixel when an error between the pixel value predicted by the prediction unit and the pixel value of the target pixel is within a predetermined threshold. You may make it judge that it is.
[0033]
According to the above configuration, as well as encoding the identification number of the prediction unit in which the prediction is correct among the plurality of prediction units as in the related art, the plurality of prediction units perform the same prediction, and Focusing on the occurrence of codewords that are not used when the prediction is out of order, the codeword can be reassigned to reduce this redundancy, so that the compression rate can be further increased.
[0034]
DETAILED DESCRIPTION OF THE INVENTION
Hereinafter, the concept of the present invention will be described. First, the premise of the description will be described.
[0035]
In the following description, the predictor having the configuration shown in FIG. 7A is used, and the pixel values of the peripheral pixels a, b, c, and d of the target pixel X are used as predicted values, as in the prior art section. The predictive hit rate is assumed to be a>b>c>d> (probability that all predictions are lost), and the identification numbers are 0, 1, 2,. . . Shall be assigned. However, these are assumptions for explanation, and do not limit the present invention.
[0036]
Based on the above assumption, the Huffman code tree shown in FIG. 7C is constructed for the identification numbers arranged in the order of occurrence probability, and the correspondence table of the identification numbers and variable-length codewords shown in FIG. 7B is obtained. .
[0037]
FIG. 2 is a diagram illustrating the state of the pixels in the predictor.
[0038]
As shown in FIG. 2A, when the reference pixels b and c and the target pixel X have the same color (vertical line region), the reference pixel b matches the target pixel X, and therefore, the 2-bit code word “10”. Is output. Although the target pixel c has the same color, the code word output when this pixel is adopted is “110” of 3 bits, so that it is not selected for the purpose of encoding.
[0039]
FIG. 2B shows a reference pixel when decoding the pixel of interest X in FIG. When decoding the pixel of interest X, if the reference pixels b and c are the same color (vertical line area), the reference pixel may be the code word “10” indicating the same color as the reference pixel b. The code word “110” indicating the same color as c can be excluded for the reason described above.
[0040]
As shown in FIG. 2C, the reference pixels b and c are the same color (vertical line area), and the reference pixel d and the target pixel X are the same color (diagonal line area) different from the reference pixels b and c. ).
[0041]
In this case, a 4-bit code word “1110” indicating a match with the reference pixel d is output. However, as described above, since the code word “110” indicating the match with the reference pixel c is not used, it can be temporarily reassigned as a code indicating the match with the reference pixel d. . In decoding, 1-bit information is reduced by making the same arrangement.
[0042]
FIG. 3 shows combinations in which bits can be reduced when four predictors (reference pixels) are used. The table in FIG. 7B is used for the correspondence between the predictor and the code word.
[0043]
When the prediction of the predictor of the first column in FIG. 3 is correct and the prediction errors of the predictors shown in the second column match, the codeword of the third column can be changed. It represents the number of bits to be reduced.
[0044]
For example, when the predictor 2 matches the target pixel, when the predictors 0 and 1 (a and b) match in colors different from these, the code “10” assigned to the predictor 1 is Since it is not used, it can be used instead of the code word “110” assigned to the predictor 2.
[0045]
FIG. 4 shows the results of measuring how much the case where the conditions shown in FIG. 3 are satisfied occurs in an actual image. For the measurement, two types of CG images described in PDL were used. Image A is a geometric pattern composed of straight lines, and image B is an LSI wiring diagram. The number of pixels is 3312 × 4680, and the K component of the color space YMCK is used.
[0046]
The code amount reduced from the sum of products of the number of bits reduced under each condition shown in FIG. 3 and the number of occurrences of each condition in images A and B shown in FIG. 4 can be obtained. In the case of image A, the code amount of 341,080 bits (42,635 bytes) is reduced, and in the case of image B, the code amount of 669,427 bits (83,678 bytes) is reduced.
[0047]
【Example】
A first embodiment of the present invention will be described.
[0048]
FIG. 1 is a block diagram for explaining the configuration of the first embodiment of the present invention. In addition, about the structure same as the structure of the prior art shown by FIG. 8, the same number is attached | subjected and description is abbreviate | omitted.
[0049]
In the same figure, 410 detects a state corresponding to the condition of FIG. 3 based on the prediction error signal output from the subtraction units 401, 402, and 403 and the correspondence information between the prediction unit and the identification number in the identification number determination unit 408. And a determination unit that outputs the code change information to the encoding unit 50.
[0050]
Next, the operation will be described based on the flowchart of FIG.
[0051]
First, the determination unit 410 detects an identification number output by the identification number determination unit 408, that is, a predictor that performs prediction that matches the target pixel and satisfies the condition of the first column in the table of FIG. 3 (FIG. 5). -S1).
[0052]
Next, the determination unit 410 obtains correspondence information between the prediction unit and the identification number from the identification number determination unit 408, and determines the second column of the table in FIG. 3, “a combination of predictors for comparing predicted values” ( FIG. 5-S2).
[0053]
Further, the determination unit 410 reads the prediction errors of the combination prediction units determined in FIG. 5-S2 from the prediction errors output from the subtraction units 401, 402, 403, respectively (FIG. 5-S3), and determines whether they match. Determination is made (FIG. 5-S4).
[0054]
If they match, the encoding unit 50 is notified of the change of the code word in the third column in the table of FIG. 3 (FIG. 5-S5).
[0055]
With the above configuration and operation, image coding is performed by the image coding apparatus of the present invention.
[0056]
Since the decoding operation is more obvious than the encoding operation, description thereof is omitted.
[0057]
The number of predictors and the prediction formula are not limited to these.
[0058]
Alternatively, the binary sequence assigned to the identification number of the correct predictor may be a variable length code, which may be multiplexed and output as code data 140, or the assigned binary sequence may be arithmetically encoded, Code data 140 may be generated as output.
[0059]
Further, when the error between the predicted pixel value and the pixel value of the target pixel is within a predetermined threshold, it may be determined that the predicted pixel value matches the pixel value of the target pixel. .
[0060]
【The invention's effect】
As described above, according to the present invention, as well as encoding a prediction unit in which a prediction is correct among a plurality of prediction units as in the prior art, a plurality of prediction units perform the same prediction, and Focusing on the occurrence of codewords that are not used when the prediction is off, re-assigning codewords can increase the compression rate.
[0061]
The specific effect of the present invention largely depends on the content of the image to be encoded and how to assign the variable length code. However, as described in the section of the embodiment, the code amount for the actual image The reduction effect is confirmed.
[Brief description of the drawings]
FIG. 1 is a block diagram showing a configuration of a first exemplary embodiment of the present invention.
FIG. 2 is an explanatory diagram of a pixel arrangement.
FIG. 3 is a diagram for explaining reassignment of codes.
FIG. 4 is a diagram for explaining an occurrence frequency in an actual image.
FIG. 5 is a flowchart for explaining the operation;
FIG. 6 is a diagram illustrating a data configuration.
FIG. 7 is an explanatory diagram of encoding.
FIG. 8 is a block diagram showing a configuration of a conventional technique.
[Explanation of symbols]
10: Image input unit,
20 ... 1st prediction part,
21 ... 2nd prediction part,
22 ... Third prediction unit,
30 ... Prediction error calculation unit,
40 ... selection part,
50: Encoding unit,
51. Decoding unit,
60: Code output unit,
100: image data,
110 ... predicted value data,
120 ... prediction error data,
130 ... Predicted situation data,
140: Code data,
401, 402.403 ... subtractor,
404, 405, 406 ... Zero determination unit,
407 ... AND,
408 ... identification number determination unit,
409 ... multiplexing unit,
410 ... determination unit

Claims (3)

An image input means for inputting an image, a plurality of prediction means for predicting the pixel value of the target pixel to be encoded in the image input by the image input means by different prediction methods, and input by the image input means A prediction error calculation unit that calculates an error between a pixel value of a target pixel in the image and a predicted value predicted by a predetermined prediction method, and a pixel value predicted by the plurality of prediction units and a pixel value of the target pixel match When the pixel value predicted by any one of the plurality of prediction units matches the pixel value of the target pixel by the match determination unit that determines whether or not to match, The prediction means determined to match is output, and if it is not determined to match, the selection means for selecting and outputting the error, and the identification information and error output by the selection means are encoded. In the image encoding apparatus having coding means, and output means for outputting the encoded code by the encoding means of,
That at least one prediction result of the plurality of prediction means matches the pixel value of the target pixel, and the predicted pixel values of at least two prediction means determined to be inconsistent with the pixel value of the target pixel are equal. Determining means for determining to change the assignment of the binary sequence in the encoding means based on the length of the variable length binary sequence given to the identification information corresponding to each prediction means in the encoding means With
In the encoding means, among the plurality of prediction means, a variable length binary sequence of length k (k is a natural number) is included in the identification information of the prediction means in which the predicted pixel value and the pixel value of the target pixel match. When assigned
Among the plurality of prediction means, there are a plurality of the same pixel values predicted by the prediction means in which the predicted pixel value does not match the pixel value of the target pixel, and the identification information of these prediction means is When a binary sequence shorter than length k is allocated in the encoding means,
From the determination means that the encoding means determines to change the assignment of the binary sequence in the encoding means based on the length of the variable length binary sequence given to the identification information corresponding to each prediction means. When the binary sequence allocation change notification is received, the length of the binary sequence assigned to the identification information of the prediction means in which the predicted pixel value matches the pixel value of the target pixel is the same as that before the binary sequence assignment change. A plurality of binary sequences shorter than the length k assigned to a part of the identification information of each of the plurality of prediction means having the same predicted pixel value so as to be shorter than the length are equal to each other. An image encoding apparatus characterized by assigning to identification information of other prediction means .   The image encoding apparatus according to claim 1, wherein the variable length binary sequence is a Huffman code. The coincidence determining unit matches the predicted pixel value with the pixel value of the target pixel when an error between the pixel value predicted by the prediction unit and the pixel value of the target pixel is within a predetermined threshold. the image coding apparatus according to claim 1 or 2 determines that.

Images