JP2007243306A - Image coder and method thereof, and image decoder and method thereof - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To utilize advantages of both coding using irreversible frequency conversion and reversible encoding using a predictor. <P>SOLUTION: In coding, a source coder 101 predicts an input image per pixel using the predictor, and transmits an identifier and a prediction error as a result thereof. A frequency converter 102 uses its wavelet transform (DWT) to apply DWT processing to the prediction error transmitted from the source coder 101, and transmits it. A quantizer of the frequency converter 102 performs quantization processing to pixel values transmitted from the DWT, and transmits the values. An EBCOT 103 merges both the information in a depthwise direction in a way that a quantization index value subjected to the quantization processing for each pixel value is arranged on an LSB side, and identification information is arranged on an MSB side; executes EBCOT processing; and outputs a resulted code. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image coding technique, and in particular, aims to provide a coding method having both the advantages of transform coding and the advantages of predictive coding.

  An international standard for image coding, JPEG 2000, is known. JPEG2000 can obtain a high compression rate by performing irreversible by discrete wavelet transform (DWT) and quantization on a normal natural image. By the way, depending on the image, it is possible to obtain a higher performance than the lossless encoding performance of JPEG2000 alone by putting a prediction result such as JPEG-LS into an entropy encoder of JPEG2000. This method is advantageous in the case of a low noise image such as that generated by a computer.

  (1) Predictive coding + JPEG2000 entropy coding (reversible coding) and (2) JPEG2000 irreversible coding are selected and coded, so that they do not depend on the type of input image. In addition, high encoding efficiency can be obtained for any input image.

  A conventional example in which encoding is performed by switching between lossless encoding and lossy encoding will be described below.

  Patent Document 1 includes an irreversible type encoding unit and a lossless type encoding unit. When the number of gradation levels of multilevel image information is large, the irreversible type encoding unit is used and the number of gradation levels is small. In this case, lossless encoding means is used.

  Patent Document 2 proposes a method of performing compression with a plurality of encoding methods for each page and selecting the one with the highest compression rate.

  Patent Document 3 proposes a method of switching between reversible / irreversible by looking at the number of colors for each block (for example, 4 × 4 pixels).

  Patent Document 4 proposes switching in block units, measuring the number of pixels having the same value as the adjacent pixel in the block, and performing lossless encoding when the number of pixels is large. .

  Patent Document 5 proposes switching between reversible / irreversible for each pixel. Lossy encoding is DPCM, and lossless encoding is dictionary-based, and raster scanning is performed for both lossless and lossy encoding. When dictionary registered pixel values exist or when the same pixel values are continuous, the mode is changed to the reversible mode. When the same pixel value does not exist in the surrounding area, the mode is changed to the irreversible mode.

  In addition, the present applicant has proposed an image encoding method as shown in FIG. 1 (currently unpublished. Japanese Patent Application No. 2005, please describe the application number of FE04-06575).

  In FIG. 1, an encoding unit 10 includes a source coder (predictive encoder) 11 and a JPEG2000 encoding unit 12, and a decoding unit 20 includes a source decoder (inverse prediction unit) 21 and a JPEG2000 decoding unit 22, and compression / decompression. The control unit 30 switches between prediction encoding / decoding and JPEG encoding / decoding. In the example shown in the drawing, it is assumed that the image encoded by the encoding unit 10 is temporarily stored in a storage device (not shown), and then decoded and output.

When the input image is a character / graphics image, a signal for switching to a scheme using a source coder instead of JPEG2000 is supplied to the compression / decompression control unit 30. On the other hand, when the input image is mainly a natural image, a signal for switching to a scheme using JPEG 2000 is supplied to the compression / decompression control unit 30. When the switching signal is JPEG2000, the compression / decompression control unit 30 transmits a select signal that bypasses the source coder 11 and the source decoder 21, and the bypass instruction signal is sent to the JPEG2000 encoding unit 12 and the JPEG2000 decoding unit 22. Do not send. Conversely, when the switching signal uses the scheme of JPEG2000 to source coder, the select signal for selecting the source coder 11 and the source decoder 21, and the bypass instruction of DWT + Q and IDWT + Q ^{−1 to} the JPEG2000 encoder 22 and JPEG2000 decoder, respectively. Send a signal.

  According to this proposal, an encoding mode can be selected adaptively for each input image.

  However, the proposals in Patent Documents 1 and 2 and the proposal shown in FIG. 1 are systems that switch between reversible and irreversible for each page, and support documents in which pixels suitable for reversibility and pixels suitable for irreversibility coexist. I can't.

  The proposals in Patent Documents 3 and 4 are a method of switching between reversible / irreversible for each block, and cannot cope with a case where an image suitable for reversibility and an image suitable for lossy are mixed in a block. .

The proposal of Patent Document 5 is a method of switching for each pixel, and can cope with mixed documents. However, the lossy compression method uses raster scan order DPCM, and the compression efficiency is lower than that of transform coding generally adopted as the lossy compression method. That is, JPEG2000 cannot be adopted.
JP-A-6-261215 Japanese Patent Laid-Open No. 10-75370 JP 2000-333017 A JP 2001-43363 A JP 2001-257888 A

  The present invention has been made in consideration of the above circumstances. Regardless of the nature of the input image, JPEG2000 has excellent lossy compression performance and excellent compression performance of both lossless encoding using a predictor. An image encoding method that holds the image is to be provided.

  According to the principle configuration example of the present invention, in order to achieve the above-described object, the image coding apparatus: means for inputting an image; and separating a pixel value from which a value can be accurately predicted and a pixel value from which the value is not predicted And means for outputting information that is sufficient to restore a pixel value that can be accurately predicted, and information that indicates the position of a pixel value that can be accurately predicted and a pixel value that cannot be accurately predicted; Means for irreversibly frequency-converting pixel value errors; and means for entropy coding by merging information sufficient to restore the accurately predictable pixel values and the irreversible frequency-transformed coefficients. Is provided.

  In this configuration example, it is possible to achieve both excellent points of encoding by irreversible frequency conversion and lossless encoding using a predictor.

  Further, the present invention will be described.

  According to the present invention, in order to achieve the above-described object, the image coding apparatus includes: an input unit that inputs an image; a prediction unit that predicts a pixel value; and between a pixel value and a prediction result of the prediction unit Calculation means for calculating the error of the output; output means for outputting information for restoring the prediction result of the prediction means; irreversible frequency conversion means for outputting coefficient information by applying irreversible frequency conversion to the error And encoding means for connecting the information for restoring the prediction result and the coefficient information of the frequency conversion means to entropy-encode the connection result.

  In this configuration, it is possible to achieve both excellent points of encoding by irreversible frequency conversion and lossless encoding using a predictor.

  In this configuration, the information for restoring the prediction result of the prediction unit may be an identifier of one or a plurality of predictors constituting the prediction unit, or may be the order of the predictors. The order is, for example, the order of descending prediction hits.

  Further, in the preceding stage of the irreversible frequency conversion, appropriate pixel values are given to peripheral pixels of pixels that include errors and the values cannot be accurately predicted, and the coefficient information generated by the irreversible frequency conversion is reduced. You may further have a means. In this case, typically, an appropriate pixel value is the result of low-pass filtering after setting the value of the surrounding pixels of a pixel whose value cannot be accurately predicted to 0.

  In addition, the encoding unit typically arranges the information for restoring the prediction result on the MSB side and the coefficient information output from the irreversible frequency conversion unit on the LSB side and connects the information to restore the prediction result. The information for restoring the result is reversibly encoded, and the coefficient information is irreversibly encoded by code truncation.

  According to the present invention, the image decoding apparatus for decoding the code output from the image encoding apparatus described above: means for inputting the code; means for entropy decoding the input code; decoding result of entropy decoding Means for calculating an error by applying an irreversible inverse frequency transform to the coefficient information of: a means for restoring a prediction result from information for restoring a prediction result of decoding results of entropy decoding; the error and the above Means for generating a decoded image from the prediction result.

  In this configuration, it is possible to decode a code that combines the advantages of both irreversible frequency conversion encoding and lossless encoding using a predictor.

  The present invention can be realized not only as an apparatus or a system but also as a method. Of course, a part of the invention can be configured as software. Of course, software products used to cause a computer to execute such software are also included in the technical scope of the present invention.

  These and other aspects of the invention are set forth in the appended claims and will be described in detail below with reference to examples.

  According to the present invention, it is possible to perform encoding / decoding that combines the advantages of both irreversible frequency conversion encoding and lossless encoding using a predictor.

  Examples of the present invention will be described below.

  FIG. 2 shows the configuration of the image coding apparatus according to the first embodiment of the present invention. In this figure, the image coding apparatus 100 includes a source coder 101, an irreversible frequency conversion unit 102, an EBCOT (Embedded Block Coding with). An optimized truncation) unit 103 is included. The source coder 101 predicts a pixel value using one or a plurality of predictors, and an error (prediction error) between the predictor identifier (which may be the order of prediction likelihood) and the predicted value. Is output. Typically, the predictor may use a neighboring pixel value as a predicted value. The identifier of the predictor is supplied to the EBCOT unit 103, and the prediction error is converted into coefficient information by, for example, wavelet transform through the irreversible frequency conversion unit 102 and supplied to the EBCOT unit 103. The EBCOT unit 103 performs entropy coding by bit modeling and arithmetic coding. The EBCOT unit 102 is the same as that of JPEG2000.

  In this embodiment, the following processing is performed.

(1) The predictor of the source coder 101 categorizes the predictor identifier and the prediction error for each pixel.
(2) The prediction error value is converted into a frequency coefficient index by irreversible frequency conversion.
(3) The predictor identifier (for example, 3 bits) is merged with the MSB side and the frequency coefficient index (for example, 11 bits) is merged with the LSB side, and entropy coding is performed by the JPEG2000 entropy coder EBCOT unit 103. At this time, the code truncation of the EBCOT unit 103 is limited to the frequency coefficient index.

  FIG. 3 shows the configuration of the image decoding apparatus according to the first embodiment of the present invention. In this figure, the image decoding apparatus 200 includes an EBCOT unit 201, an inverse frequency conversion unit 202, and a source decoder 203. Yes. The EBCOT unit 102 entropy-decodes the code and outputs a predictor identifier and coefficient information. The inverse frequency transform unit 202 performs, for example, inverse wavelet transform from the coefficient information to generate a prediction error. The source decoder 203 decodes an image from the predictor identifier and the prediction error.

  The operations of the image encoding device 100 and the image decoding device 200 according to the first embodiment will be further described with reference to FIGS.

  First, at the time of encoding (FIG. 2), the source coder 101 predicts an input image for each pixel using a predictor, and sends an identifier and a prediction error as a result. The frequency conversion unit 102 performs DWT processing on the prediction error transmitted from the source coder 101 and transmits the result by the wavelet conversion unit (DWT). The quantization unit of the frequency conversion unit 102 performs a quantization process on the pixel value transmitted from the DWT unit and transmits the result. The EBCOT unit 103 merges both pieces of information in the depth direction in such a manner that the quantization index value quantized for each pixel value is placed on the LSB side and the identifier information is placed on the MSB side, performs EBCOT processing, and encodes the result Is sent out. This merge mode is schematically shown in FIG.

  At the time of decoding (FIG. 3), the EBCOT unit 201 performs entropy decoding on the code, and sends the LSB side prediction error quantization index and the MSB side identifier, respectively. The inverse quantization unit of the inverse frequency transform unit 202 performs an inverse quantization process on the quantization index of the prediction error from the EBCOT unit 201 and sends the result to the IDWT unit of the inverse frequency transform unit 202. The IDWT unit sends the prediction error result after the IDWT process to the source decoder 203. The source decoder 203 performs prediction decoding / prediction error decoding based on the identifier information from the EBCOT unit 201 and the prediction error from the IDWT unit of the inverse frequency transform unit 202, generates a decoded image, and outputs the result.

  Next, an image coding apparatus 300 according to Embodiment 2 of the present invention will be described. The image decoding apparatus is the same as that in the first embodiment.

  FIG. 5 shows a configuration of an image encoding device 300 according to the second embodiment. In this figure, the image encoding device 300 includes a source coder 101, an irreversible frequency conversion unit 102, an EBCOT unit 103, and a hole filling unit 301. It is configured to include. In FIG. 5, parts corresponding to those in FIG.

  Also in this example, the source coder 101 outputs a predictor identifier and a prediction error with respect to the pixel value of the input image. As illustrated in FIG. 6, the hole filling unit 301 performs a hole filling process on the surrounding pixels of the prediction error pixel using, for example, a low-pass filter, and transmits the result. The DWT unit of the irreversible frequency conversion unit 102 performs DWT processing on the prediction error after filter processing sent from the hole filling unit 301 and sends coefficient information. The quantization unit of the irreversible frequency conversion unit 102 performs a quantization process on the pixel value transmitted from the DWT unit, and transmits coefficient information after the quantization process. The EBCOT unit 103 merges both pieces of information in the depth direction in such a manner that the quantization index value quantized for each pixel value is arranged on the LSB side and the identifier information on the MSB side, performs EBCOT processing, and encodes the result Is sent out.

  In this embodiment, since the change in the prediction error becomes gentle, the coefficient information can be reduced. As long as the information amount is reduced, the hole filling unit 301 may not be a low-pass filter, and may be an average value process, for example.

  Next, a third embodiment in which the present invention is applied to a printer will be described.

  FIG. 7 shows the entire printer 400 of the third embodiment. In FIG. 7, portions corresponding to those in FIG. 2, FIG. 3, and FIG.

  In FIG. 7, the printer 400 rasterizes a document described in a page description language, encodes the rasterized image, stores it in the hard disk 402, decodes it, and outputs it to the image forming apparatus 403. . The image forming apparatus 403 forms an image by, for example, an electrostatic photography method.

  In FIG. 7, the printer 400 includes a decomposer 401, an image encoding device 300, an image decoding device 200, a hard disk 402, and an image forming device 403. In this example, the DWT unit 102a and the quantization unit 102b constituting the frequency conversion unit 102 are explicitly shown. Further, the inverse quantization unit 202b and the IDWT unit 202a constituting the inverse frequency conversion unit 202 are explicitly shown. Note that the functional blocks in this example can be configured as software. For example, the computer program 404 can be installed in a computer constituting the printer 400 and the hardware resources and software resources of the computer can be cooperated.

  The configurations and operations of the image encoding device 300 and the image decoding device 200 are as described above. The image encoding device 100 may be adopted instead of the image encoding device 300.

  In the printer 400 of this example, a code can be stored in the hard disk 402 until it is expanded in synchronization with the image forming apparatus 403, and is expanded and sent out in synchronization with the image forming apparatus 403.

  The present invention is not limited to the above-described embodiments, and various modifications can be made without departing from the spirit of the invention.

It is a block diagram explaining the background art of this invention. It is a block diagram which shows the structure of the image coding apparatus of Example 1 of this invention. It is a block diagram which shows the structure of the image decoding apparatus of Example 1 of this invention. It is a figure which shows typically the aspect which merges the predictor identifier information of the above-mentioned Example and the coefficient information of a prediction error. It is a block diagram which shows the structure of the image coding apparatus of Example 2 of this invention. It is a figure which shows typically the hole-filling process of the said Example 2. FIG. It is a block diagram which shows the structure of Example 3 which applied this invention to the printer.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Image coding apparatus 101 Source coder 102 Non-reversible frequency transformation part 102a DWT part 102b Quantization part 103 EBCOT part 200 Image decoding apparatus 201 EBCOT part 202 Inverse frequency transformation part 202b Inverse quantization part 202a IDWT part 203 Source decoder 300 Image coding Apparatus 301 hole filling unit 400 printer 401 decomposer 402 hard disk 403 image forming apparatus 404 computer program

Claims (11)

An input means for inputting an image;
Prediction means for predicting pixel values;
Calculation means for calculating an error between the pixel value and the prediction result of the prediction means;
Output means for outputting information for restoring the prediction result of the prediction means;
Irreversible frequency transforming means for applying coefficient transform to the error and outputting coefficient information;
An image encoding apparatus comprising: encoding means for concatenating the information for restoring the prediction result and the coefficient information of the frequency conversion means to entropy-encode the connection result.   The image encoding apparatus according to claim 1, wherein the information for restoring the prediction result of the prediction means is an identifier of one or more predictors constituting the prediction means.   The image coding apparatus according to claim 1, wherein the information for restoring the prediction result of the prediction means is an order of one or more predictors constituting the prediction means.   Means for reducing the coefficient information generated by the irreversible frequency conversion in the preceding stage of the irreversible frequency conversion, by giving a pixel value of an appropriate value to the peripheral pixels of the pixel that includes an error and whose value cannot be accurately predicted. Furthermore, the image coding apparatus in any one of Claims 1-3.   5. The image encoding device according to claim 4, wherein the appropriate pixel value is a result obtained by setting a value of a peripheral pixel of a pixel whose value cannot be accurately predicted to 0 and further performing a low-pass filter process.   The encoding unit arranges and connects the information for restoring the prediction result on the MSB side and the coefficient information output from the irreversible frequency conversion unit on the LSB side, and the information for restoring the prediction result is reversible. The image coding apparatus according to claim 1, wherein the coefficient information is also entropy-encoded irreversibly by code truncation. In the image decoding apparatus which decodes the code | symbol output from the image coding apparatus of Claim 1,
Means for inputting the code;
Means for entropy decoding the input code;
Means for applying an irreversible inverse frequency transform to coefficient information in the decoding result of entropy decoding and calculating an error;
Means for restoring the prediction result from information for restoring the prediction result of the decoding result of entropy decoding;
An image decoding apparatus comprising: means for generating a decoded image from the error and the prediction result. Inputting an image;
Predicting pixel values;
Calculating an error between the pixel value and the prediction result of the prediction means;
Outputting information for restoring the prediction result of the prediction means;
Applying coefficient of irreversible frequency transformation to the error and outputting coefficient information;
An image encoding method comprising the step of concatenating the information for restoring the prediction result and the coefficient information of the frequency conversion means and entropy encoding the concatenation result. In the image decoding method which decodes the code | symbol produced | generated by the image coding method of Claim 8,
Inputting the above sign;
Entropy decoding the input code;
Applying an irreversible inverse frequency transform to the coefficient information of the decoding result of entropy decoding to calculate an error;
Restoring the prediction result from the information for restoring the prediction result of the decoding result of the entropy decoding; and
And a step of generating a decoded image from the error and the prediction result. Inputting an image;
Predicting pixel values;
Calculating an error between the pixel value and the prediction result of the prediction means;
Outputting information for restoring the prediction result of the prediction means;
Applying coefficient of irreversible frequency transformation to the error and outputting coefficient information;
A computer program for image encoding, which is used for causing a computer to execute the step of entropy encoding the concatenation result by concatenating the information for restoring the prediction result and the coefficient information of the frequency conversion means. An image decoding computer program used by a computer to decode a code generated by the image encoding method of claim 8,
Inputting the above sign;
Entropy decoding the input code;
Applying an irreversible inverse frequency transform to the coefficient information of the decoding result of entropy decoding to calculate an error;
Restoring the prediction result from the information for restoring the prediction result of the decoding result of the entropy decoding; and
An image decoding computer program used for causing a computer to execute a step of generating a decoded image from the error and the prediction result.