JP2007028531A - Image coding apparatus and image coding method - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide an image coding apparatus and an image coding method capable of excellently coding an images of a character and a graphic and executing code amount control at a high speed without the need for repetitive processing. <P>SOLUTION: A prediction section 1 applies prediction processing to each pixel of a received image and gives the result of prediction to a bit plane coding section 2. The bit plane coding section 2 decomposes a received prediction result into a bit plane and sequentially codes each bit plane from a high-order bit plane to a low-order bit plane. In this case, the bit plane coding section 2 integrates a code amount for each sequential coding from the high-order bit plane and compares the integrated code amount with a target code amount. When the integrated code amount exceeds the target code amount, the coding processing after the bit plane is terminated and immediately preceding code data are outputted. Thus, the code amount control can be conducted by one path in spite of the prediction coding. <P>COPYRIGHT: (C)2007,JPO&INPIT

Description

  The present invention relates to an image coding technique for coding an image by predictive coding, and particularly to an image coding technique capable of controlling a code amount in predictive coding.

  As a technique for encoding and decoding still images, a JPEG (Joint Photographic Expert Group) method has been widely used. Recently, JPEG2000 (hereinafter abbreviated as J2K) has been newly standardized. FIG. 4 is a block diagram showing an outline of the encoding process in J2K. In the figure, 21 is a pre-processing unit, 22 is a wavelet transform unit, 23 is a quantization unit, 24 is an EBCOT unit, and 25 is a post-processing unit. The preprocessing unit 21 appropriately performs processing such as DC level shift, color conversion, and tile division, and the wavelet conversion unit 22 performs discrete wavelet conversion. The coefficient generated by the wavelet transform is quantized by the quantization unit 23. The processing of the quantization unit 23 is optional. Thereafter, the EBCOT unit 24 performs bit modeling, entropy coding by arithmetic coding, and rate control. This EBCOT (Embedded Block Coding with Optimized Truncation) is also a feature of J2K. Finally, processing such as bit stream generation is performed by the post-processing unit 25 to obtain an encoded image.

  Unlike JPEG, J2K seamlessly realizes lossless encoding and lossy encoding, and realizes high image quality and a high compression rate. However, as described above, since J2K performs wavelet transform, which is one of orthogonal transform coding, in principle, it is not suitable for compression in which many edges included in characters and graphics images are stored, and lossless compression performance is poor. There is a problem.

  As described above, since orthogonal transform coding is not suitable for compression of characters and graphics images, for example, in Patent Document 1, lossless coding is performed on a block in which characters exist to improve reproducibility of characters and the like. I am letting. With lossless encoding, character edges are preserved and reproduced well during decoding. However, lossless encoding has a drawback that the amount of code cannot be controlled.

  As a coding method suitable for compression of characters and graphics images, there is a predictive coding method. The predictive coding method is a method of predicting a value of a target pixel from surrounding pixels (context), and setting a difference between the predicted value and the target pixel as a prediction result and encoding the prediction result. For example, if the prediction value always matches the target pixel, the prediction results are all 0, so that highly efficient encoding can be performed. In general, when the prediction result takes a small value, the code amount becomes small. Therefore, the code amount can be controlled by changing a parameter at the time of prediction.

  As a method of controlling the code amount in the conventional predictive coding method, encoding is performed with a certain prediction parameter, it is determined whether or not the obtained code amount is equal to or less than the target code amount, and exceeds the target code amount. In this case, the encoding is repeated many times such as changing the prediction parameter and encoding again. However, repeating such an encoding process takes a long time, so improvement is desired.

  In J2K described above, code amount control can be performed by rate control during EBCOT, and code data that satisfies a target code amount can be obtained by a single encoding process. However, as described above, the compression rate of characters and graphic images decreases due to the influence of wavelet transform and the like, so that there is a problem that the image quality of characters and graphic images deteriorates when rate control is performed.

  As a technique before the above-described EBCOT, it is also considered that an image is decomposed into bit planes and compressed using a binary coding method such as JBIG for each bit plane. However, with this method, although lossless encoding is possible, the compression rate is lower than that of a general predictive encoding method. Further, although the code amount can be controlled by stopping the encoding of the lower bit plane, there is a problem that the image quality is greatly affected by the control of the code amount because the compression rate is low.

  As described above, in the conventional encoding method, there is no encoding method that suppresses deterioration of image quality for characters and graphic images and encodes at a high compression rate, and development of such an encoding method has been awaited. .

JP-A-6-152982

  The present invention has been made in view of the above-described circumstances, and is capable of encoding a character or graphics image satisfactorily, and an image encoding apparatus and image capable of performing code amount control at high speed without repeating processing. The object is to provide an encoding method.

  The present invention is characterized in that a prediction process is performed on each pixel of an input image, and the prediction result is bit-plane encoded from the upper bit plane toward the lower bit plane until the target code amount is reached. . When a code identifier is output together with a prediction error during the prediction process, the code identifier output by the prediction process is the upper bit plane, the prediction error is the lower bit plane, and the upper bit plane is changed to the lower bit plane. The encoding is preferably performed until the target code amount is reached, and the encoding process is not terminated in the bit plane of the code identifier. Bit plane encoding can be performed using, for example, JPEG2000 EBCOT. The prediction process can use a JPEG-LS predictor or the like.

  According to the present invention, prediction processing of a predictive encoding scheme that is optimal for encoding a character / graphics image is performed, and the prediction result is divided into bit planes to perform bit plane encoding. This makes it possible to significantly reduce the amount of code compared to the case of simply performing bit-plane coding on an image, and performs coding processing that preserves the edges of characters and graphics images while maintaining a small amount of code. be able to. Furthermore, code amount control is enabled by performing encoding from the upper bit plane toward the lower bit plane until the target code amount is reached during bit plane encoding. At this time, by encoding from the upper bit plane that is important for the image, even if the encoding is terminated due to exceeding the target code amount, the coding for the lower bit plane that is less important is terminated. Can be suppressed. Furthermore, since the amount of code is reduced by performing the prediction process in the previous stage, even when encoding is terminated with the same target code amount, encoding is performed up to the lower bit plane compared to the case of simply performing bit plane encoding. Therefore, deterioration in image quality can be suppressed.

  Further, in the conventional predictive encoding process, the iterative process by changing the parameters is performed for the code amount control. However, in the present invention, the encoding process can be completed in a so-called one pass without performing the iterative process. . Accordingly, it is possible to obtain code data of the target code amount at a higher speed than in the predictive encoding process with conventional code amount control.

  As described above, according to the present invention, it is possible to perform encoding processing in which edges are preserved for characters and graphics images, to achieve both a high compression rate and a code amount control function, and to perform encoding processing at high speed. It has the effect of becoming.

  FIG. 1 is a block diagram showing an embodiment of the present invention. In the figure, 1 is a prediction unit, and 2 is a bit plane encoding unit. The prediction unit 1 performs a prediction process for each pixel of the input image. The prediction method is arbitrary. As the prediction unit 11, for example, a JPEG-LS predictor, a source coder of an existing predictive coding method, such as PNG or DPCM, can be applied.

  The bit plane encoding unit 2 decomposes the prediction result output from the prediction unit 1 into bit planes, and encodes each bit plane. At this time, encoding is performed for each bit plane in order from the upper bit plane to the lower bit plane. FIG. 2 is an explanatory diagram of a bit plane. In FIG. 2, the predicted value (binary representation) is shown in the vertical direction, and in the figure, the upper side is the upper bit (MSB) and the lower side is the lower bit (LSB). Such a prediction value is decomposed for each bit, and, for example, the most significant bit plane is generated by extracting the most significant bit for the prediction value corresponding to each pixel. Similarly, for the predicted value corresponding to each pixel, the second bit plane from the most significant bit is generated by taking out the second bit from the most significant bit. In this way, a plurality of bit planes are generated by extracting bits at the same position from the prediction values corresponding to the respective pixels. The example described in FIG. 2 is basic, but is not limited to this example, and the decomposition method into bit planes is arbitrary.

  Then, encoding is performed in order from the upper bit plane. In the example shown in FIG. 2, since a bit plane is generated for each bit, a binary encoding method can be used for encoding.

  Further, the bit plane encoding unit 2 receives an instruction of the target code amount from the outside, integrates the code amount every time it sequentially encodes from the upper bit plane, and compares it with the target code amount. When the integrated code amount exceeds the target code amount, the encoding process after the bit plane is terminated, and the code data up to immediately before is output. As a result, code data within the range of the target code amount can be acquired in one pass without performing iterative processing. It should be noted that code data in the case of all 0s (or all 1s) may be appropriately added to bit planes for which encoding processing has been terminated.

  In general, a prediction error is output as a prediction result, but the magnitude of the prediction error often indicates the magnitude of the change in the image, and the higher order bits of the prediction error have a greater influence on the image quality than the lower order bits. Therefore, by encoding from an important upper bit plane, the influence when the encoding process is terminated is reduced. Also, the compression rate is higher when the prediction result is bit-plane encoded than when the image is directly bit-plane encoded, and the amount of code in each bit plane can be reduced. Therefore, if the target code amount is the same, the prediction result can be encoded up to the lower bit plane by encoding the prediction result rather than the bit plane encoding of the image as it is, and the image quality at the time of decoding can be improved. Deterioration can be suppressed. Of course, even if all the bit planes are encoded, if they are within the target code amount, no truncation occurs and reversible encoding can be performed.

  When a code identifier is output from the prediction unit 1 together with a prediction error, the code identifier may be encoded in order from the bit plane of the code identifier, with the code identifier being a higher order bit than the prediction error. If the information of the code identifier is missing, there is a possibility that it cannot be decoded. Therefore, the code identifier needs to be losslessly encoded. Therefore, encoding is performed as a higher-order bit plane so that the encoding of the code identifier is not interrupted by the target code amount. Also, it is necessary to control so that the encoding of the bit plane of the code identifier is not interrupted.

  FIG. 3 is an explanatory diagram of a specific example in one embodiment of the present invention. 11 is a JPEG-LS predictor, and 12 is an EBCOT section. In the example illustrated in FIG. 3, the JPEG-LS predictor 11 is used as the prediction unit 1 illustrated in FIG. 1. Since JPEG is an irreversible encoding method and image quality deteriorates, JPEG-LS is standardized as a lossless encoding method that does not cause image quality deterioration. In this JPEG-LS, a prediction process is performed in the previous stage, and a predictor that performs this prediction process is used as the prediction unit 1 of the present invention. The JPEG-LS predictor 11 has high prediction efficiency, and can be expected to be encoded at a high compression rate in the subsequent encoding process.

  In this example, a J2K EBCOT unit 12 is used as the bit plane encoding unit 2. The EBCOT unit 12 is the EBCOT unit 24 described with reference to FIG. 4, and as described above, decomposition into bit planes by bit modeling, entropy coding by arithmetic coding for each bit plane, coding with a target code amount Each process of rate control for controlling censoring is performed. In the bit modeling process, after each block is decomposed into bit planes as shown in FIG. 2, each bit plane is decomposed and arranged into three coding passes (sub-bit planes). Each of these three coding passes is arithmetic coded. This processing is performed from the upper bit plane. By the rate control, the code data exceeding the given target code amount is discarded. Even when truncation occurs, it is possible to suppress the influence on the image quality as much as possible because the lower-order bit plane is discarded.

  In J2K, the EBCOT unit 12 performs an encoding process on the wavelet-transformed coefficient as shown in FIG. 4, but even if an encoding process is performed on the prediction result as in the present invention, the EBCOT unit 12 is highly efficient. It can be encoded with efficiency. Conversely, by performing prediction processing instead of wavelet transform, it is possible to achieve a high compression rate while preserving as many edges as possible in characters and graphics images.

  In this way, by using the JPEG-LS predictor 11 and the J2K EBCOT unit 12 in combination, it is possible to realize a high compression ratio while preserving edges existing in characters and graphics images as much as possible. Further, the code amount control with the set target code amount can be performed, and the influence on the image quality in that case can be suppressed as much as possible. Furthermore, since a high compression ratio can be realized by the high-efficiency prediction processing by the JPEG-LS predictor 11 and the bit modeling and encoding processing of the EBCOT unit 12, the conventional bit-plane encoding terminates the encoding in many bit planes. Even in such a case, the number of bit planes to be cut off can be reduced or eliminated, so that the image quality when decoded can be improved as compared with the conventional case. Furthermore, since the EBCOT unit 12 is a method that can perform code amount control in one pass, for predictive coding that has conventionally performed code amount control in an iterative process, the prediction coding process as a whole is encoded in one pass. Quantity control can be performed.

  The present invention is not limited to the specific example shown in FIG. 3, and means for performing an arbitrary prediction process is applied as the prediction unit 1, and encoding is performed for each arbitrary bit plane as the bit plane encoding unit 2. A means for performing can be applied. For example, the bit modeling of the EBCOT unit 12 may be decomposition into a simple bit plane as shown in FIG. 2, or JBIG may be used instead of arithmetic coding. In addition, as described above, when a predictive coding source coder is used in which the prediction unit 1 outputs a code identifier together with a prediction error, the code identifier is arranged on the higher bit side of the prediction error and divided into bit planes. However, bit plane encoding may be performed from the upper bit plane. In this case, it is only necessary to control the bit plane of the code identifier so that the encoding is not interrupted.

  When decoding the code data encoded according to the present invention, the encoding process may be reversed. That is, when decoding the code data encoded with the configuration shown in FIG. 1, it is decoded by the bit plane decoding means, combined for each bit to restore the prediction value, and then the inverse prediction means from the prediction value Thus, the original image (plain text) may be restored. For example, in the case of code data encoded according to the specific example shown in FIG. 3, the prediction value of the code data is restored by the inverse EBCOT means, and the reverse prediction process is performed by the JPEG-LS inverse prediction means to obtain the original image. be able to.

It is a block diagram which shows one Embodiment of this invention. It is explanatory drawing of a bit plane. It is explanatory drawing of the specific example in one Embodiment of this invention. It is a block diagram which shows the outline | summary of the encoding process in J2K.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 ... Prediction part, 2 ... Bit plane encoding part, 11 ... JPEG-LS predictor, 12 ... EBCOT part, 21 ... Pre-processing part, 22 ... Wavelet transformation part, 23 ... Quantization part, 24 ... EBCOT part, 25 ... post-processing section.

Claims (8)

  A prediction unit that performs a prediction process on each pixel of the input image, and a bit plane encoding unit that encodes the output of the prediction unit from the upper bit plane toward the lower bit plane until the target code amount is reached. An image encoding device.   The image encoding apparatus according to claim 1, wherein JPEG2000 EBCOT is used as the bit plane encoding means.   The predicting unit outputs a code identifier together with a prediction error, and the bit plane encoding unit sets the code identifier output from the predicting unit as an upper bit plane, and sets the prediction error as a lower bit plane. The image coding apparatus according to claim 1, wherein control is performed so that coding processing is not terminated in the bit plane of the code identifier.   The image encoding apparatus according to claim 1, wherein a JPEG-LS predictor is used as the predicting unit.   A prediction process is performed on each pixel of the input image by the prediction unit, and an output of the prediction process is encoded by the bit plane encoding unit from the upper bit plane toward the lower bit plane until the target code amount is reached. An image encoding method.   6. The image encoding method according to claim 5, wherein JPEG2000 EBCOT is used as the bit plane encoding means.   A code identifier is output together with a prediction error by the prediction process, the code identifier output by the prediction process is an upper bit plane, the prediction error is a lower bit plane, and a target code amount is set from an upper bit plane to a lower bit plane. The image encoding method according to claim 5 or 6, wherein the encoding is performed until the encoding reaches the bit plane, and control is performed so that the encoding process is not terminated in the bit plane of the code identifier.   The image encoding method according to claim 5 or 6, wherein a JPEG-LS predictor is used as the predictor.

Images