JPH09205647A - Image coder, image decoder and image processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To obtain an output image with high quality by suppressing a reduction in a compression efficiency arranged with addition of an identifier so as to avoid a circuit scale from being increased and cost from being increased and attaining image area separation with high accuracy and applying image processing to each image area properly. SOLUTION: An input image in which areas with different attributes such as character/line drawing, and photograph/natural image are in existence in mixture is subjected to adaptive discrete cosine transformation ADCT coding and the result is stored and outputted after decoding in an image processing unit. In the processor above, an area tag signal representing the attribute of an image area is synchronized with a DC coefficient of a pixel block and the result is subjected to entropy coding and they are separated after decoding. Thus, it is not required to add a coding circuit exclusive for identifier and a storage means.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image coding device, an image decoding device and an image processing device which employ an ADCT coding method.

[0002]

2. Description of the Related Art In recent years, a digital copying machine having an electronic sorter function and an electronic RDH (Recycle Document Handler) function for storing image data read from a document in an internal memory or a hard disk and performing sorting, filing, page editing, etc. Is becoming popular. In this type of digital copying machine, in order to store a large amount of image data, it is sufficient to increase the storage capacity such as a memory. However, since increasing the capacity leads to an increase in device scale and cost, image data is usually The method of compressing and accumulating is taken.

By the way, in the case of accumulating high quality images,
Since it is preferable to store the image data in a multi-valued state rather than in a binary state, various methods for compressing the multi-valued image data have been developed. For example ADC
T (Adaptive Discrete Cosine Transform) encoding is a lossy compression method for multivalued image data, and is known as a JPEG baseline method adopted as a standard encoding method for color facsimiles. This ADCT encoding is a compression method that is excellent in terms of both image quality and compression efficiency.

Referring now to the block diagram shown in FIG.
The process of ADCT encoding will be described. In FIG. 1, an image input from the image input device 1 is first color space converted by a color space conversion circuit 2. That is,
The color space conversion circuit 2 converts the RGB data of the input color image into YCbCr data separated into a luminance signal and a color difference signal. The image data converted in this way will be hereafter referred to as Y data of a luminance signal, Cb data of a color difference signal, and Cr.
The data are processed separately, but they all go through the same process. Therefore, in the following, the description will be focused on one signal (for example, a luminance signal).

The image data whose color space has been converted is divided into blocks in the block circuit 3 into, for example, 8 × 8 pixel blocks shown in FIG. The image data that has been divided into blocks is displayed in the DCT operation circuit 4 as shown in FIG.
It is converted into a frequency coefficient (hereinafter referred to as a DC coefficient and an AC coefficient) including a C component and an AC component. In FIG. 10, the component located in the upper left corner is the DC coefficient, which is 8 × 8.
The value is proportional to the average value of the actual image data of. Further, the other 63 components are AC coefficients, which show different frequency components, and the value thereof is proportional to the amplitude of each frequency component. As shown in this figure, when DCT processing of natural images is performed, numerical values are concentrated on low frequency components in many images,
It can be seen that only a very small value appears in the high frequency component.

Now, the DC coefficient and A thus obtained
The C coefficient is quantized by the quantization circuit 5, as shown in FIG. That is, the DC coefficient and the AC coefficient are
The value is divided by the value in the quantization table held by the quantization circuit 5 (that is, the quantization step) and rounded off. By this quantization processing, the data whose numerical value is “0” is increased in the AC coefficient in the high frequency region. That is, ADCT
In encoding, the compression efficiency is improved by increasing the amount of "0" data, but at the same time, this is also a cause of lossy upon image encoding.

FIG. 12 is a diagram showing the contents of a quantization table recommended in the JPEG baseline system which is a standard encoding system for color facsimiles. FIG. 12A shows the recommended quantization of luminance signals. A table is shown, and FIG. 6B shows a recommended quantization table for color difference signals. As can be seen from this figure, a small value is set for the low frequency component in the quantization table, considering that human visual characteristics are sensitive to low frequency components but insensitive to high frequency components. Then, a large value is set for the high frequency component.

Next, the entropy coding circuit 6 entropy codes the quantized DC coefficient and AC coefficient. Since the DC coefficient is the average density of the block, it has a correlation with the DC coefficient of the adjacent block. Therefore, in the coding of the DC coefficient, the difference between the DC coefficient of the block to be coded and the DC coefficient of the immediately preceding block is calculated, and the difference value of the DC coefficient is entropy coded.

On the other hand, the AC coefficients have relatively large values in the low frequency region, and values close to "0" in the high frequency region. Therefore, in the order shown in FIG.
The C component is zigzag-scanned and the data is rearranged so that the data of “0” is continuous. Coefficients (significant coefficients) other than “0” and “0” for this series of data strings
And the continuous length of the coefficient (insignificant coefficient) of 1 is set as a set, and this set of data is entropy coded. The above process is performed for one pixel block, and this is repeated for all blocks to encode the entire image. In this way, the encoded image data is stored in the storage device 7 and then read out in response to the output request, and the entropy decoding circuit 8 or the color space conversion circuit 12 goes through a process reverse to the encoding. As a result, it is decoded and supplied to the image output device 13.

[0010]

However, the originals handled by a digital copying machine or the like are not necessarily uniform, and depending on the originals, a text area and a photo area may coexist on the same page. When targeting such manuscripts,
There is a demand to reproduce letters more like letters and photographs more like photographs. That is, for a digital copying machine equipped with an electronic sorter function, it is required to simultaneously improve both the compression efficiency of image data and the high quality reproduction of a character area and a photo area. .

In response to such a request, for example, as shown in FIG. 14, a method of encoding the entire image without performing area separation processing, accumulating the encoded image data, decoding and outputting it is adopted. Think In this method, the entire image of input image data is processed by a single method regardless of whether it is a character area or a photograph area. Therefore, if the image processing parameter is set to a value suitable for a photograph, the character area of the decoded image is deteriorated, whereas if it is set to a value suitable for a character, the photograph area of the decoded image is deteriorated. For this reason, it is not possible to obtain high image quality for all image areas, and it is not possible to satisfy the above-mentioned requirements.

Next, for example, as shown in FIG.
Consider a method of performing area separation processing on image data after T coding. That is, in this method, the entire image is first encoded without performing the region separation process, the encoded image data is accumulated, and the read image data is read and the region separation process is performed on the decoded image data. Generate a signal. Then, when outputting an image, the area tag signal is used to perform color correction, gradation correction, screening processing, switching of filter coefficients, etc. suitable for each image area characteristic. Therefore, according to this method, it is possible to obtain a high-quality decoded image as compared with the case where the area separation processing is not performed.

However, since the ADCT coding is lossy coding, the decoded image data is somewhat deteriorated with respect to the original image data. Therefore, even if the area separation process is performed on the decoded image, the result does not always match the area of the original image. For this reason, there is a problem that a high-quality decoded image cannot be expected in a portion where the area separation accuracy is low and the area separation has an error.

Further, as shown in FIG. 16, for example, AD
Consider a method of generating a region tag signal by performing region separation processing before CT encoding processing. According to this method, the area separation accuracy is higher than that of the above method, and therefore a high quality image can be expected. However, in this method, the area tag signal generated before encoding needs to be held until the time of image output, and therefore the area tag signal also needs to be stored similarly to the image data. At this time, if the region tag signal is ADCT-encoded similarly to the image data, it deteriorates after decoding, and thus an error occurs in the region separation result. Therefore, in order to store the region tag signal, it is necessary to hold the region tag signal as it is without performing compression encoding, or to perform lossless encoding processing and store it. In this case, a device for separately encoding and storing the area tag signal is required, and the circuit scale increases correspondingly, which causes a problem of cost increase. Further, since the code amount of the entire image increases by the amount of the area tag signal added, the compression efficiency decreases.
Furthermore, when decoding image data, it is necessary to decode the area tag signal in perfect synchronization with it.
In a variable length coding method such as ADCT coding, it is difficult to synchronize the decoding timings of both because the speed at which an image is decoded is not constant. Further, when the synchronization is taken, a buffer memory is separately required, which causes an increase in circuit scale and eventually a cost increase.

The present invention has been made under such a background, and suppresses a reduction in compression efficiency due to the addition of an identifier (region tag signal) indicating an attribute of an image region, thereby increasing the circuit scale and cost. The image coding apparatus, the image decoding apparatus, and the image processing apparatus capable of avoiding the above problem, enabling high-precision area separation, and obtaining a high-quality output image by performing image processing suitable for each image area. Is intended to provide.

[0016]

In order to solve the above-mentioned problems, the invention described in claim 1 uses image data as an ADC.
In an image coding apparatus for T-coding, a block forming means for dividing input image data into blocks, an identifying means for identifying a feature of each block and generating an identifier showing the feature, and a DCT for each block. DCT transforming means for transforming to generate a DC coefficient and an AC coefficient;
Quantizing means for quantizing C and AC coefficients, synthesizing means for synthesizing the quantized DC coefficient and the identifier to generate a synthetic coefficient, adjoining the quantized AC coefficient and the synthetic coefficient It is characterized by comprising an encoding means for entropy-encoding the difference value with respect to the block and generating encoded data.

Further, the invention described in claim 2 is the first invention.
An image decoding apparatus for decoding coded data generated by the image coding apparatus according to claim 1, wherein the decoding means decodes the coded data and restores the quantized AC coefficient and the synthesized coefficient. Separating means for separating the reconstructed composite coefficient into the quantized DC coefficient and the identifier, dequantizing means for dequantizing the quantized AC coefficient and DC coefficient, and the dequantizing DC
Inverse DCT conversion means for performing inverse DCT conversion on the coefficients and AC coefficients to restore the image data of each block.

Further, the invention according to claim 3 provides the invention according to claim 1.
3. The image decoding device according to claim 2, wherein the image coding device according to claim 1, a storage unit that stores the code data coded by the image coding device, and the code data read from the storage unit. It is characterized by having and.

The invention described in claim 4 is the same as the invention described in claim 3.
In the invention described in above, the synthesizing unit generates a synthesizing coefficient by adding a bit string of the identifier to the bit string of the DC coefficient, and the separating unit is a bit string of the synthesizing coefficient decoded by the decoding unit. The DC
It is characterized in that the bit string of the coefficient and the bit string of the identifier are separated.

The invention described in claim 5 is the same as the invention described in claim 3.
In the invention described in above, the quantizing means selects and quantizes a quantizing step corresponding to the feature of each block based on the identifier generated by the identifying means, and the dequantizing means performs the decoding. The dequantization step corresponding to the quantization step by the quantization means is selected and inversely quantized based on the identifier decoded by the means.

The invention described in claim 6 is the third invention.
5 to 5, the inverse DCT
The image data of each block restored by the conversion means is provided with a correction means for performing a correction process corresponding to the characteristic indicated by the identifier decoded by the decoding means.

The invention described in claim 7 is the same as the invention in claim 6
In the invention according to, the identifying means further divides the block into sub-blocks, identifies features for each sub-block to generate each identifier, and the correcting means, for the image data of each block, It is characterized in that a correction process corresponding to the feature indicated by the identifier is performed in sub-block units.

[0023]

DETAILED DESCRIPTION OF THE INVENTION Embodiments of the present invention will be described below with reference to the drawings. A: First Embodiment (1) Overall Configuration FIG. 1 is a block diagram showing the overall configuration of an image processing apparatus according to the first embodiment of the present invention. The present embodiment is configured on the assumption that a monochrome original image is input. In FIG. 1, an image input device 1001 is, for example, a scanner or the like, which reads an image of a document and converts it into digital image data. The image input device 10
Reference numeral 01 may be a scanner or the like, or an external interface or the like for taking in digital image data supplied via an external network.

Next, the raster / block circuit 1002
The input image data is divided into pixel blocks each including 8 × 8 pixels. The area tag signal generation circuit 1003
The attribute of the block is determined from the image data of 8 × 8 pixels, and the area tag signal indicating the determination result is output. The details of the area tag signal will be described later.

Next, the DCT calculation circuit 1004 converts the image data into frequency space data in block units of 8 × 8 pixels, and outputs the DC coefficient and AC coefficient of the block. The quantization circuit 1005 quantizes the DC coefficient and the AC coefficient based on the value of the quantization table stored in the rewritable memory.

Next, the area tag signal synthesis circuit 1008
The area tag signal output from the area tag signal generation circuit 1003 and the DC coefficient quantized by the quantization circuit 1005 are combined, and the combined DC coefficient is output. The process of combining the area tag signals will be described later. Further, the entropy coding circuit 1009 performs Huffman coding on the combined DC coefficient output from the area tag signal combining circuit 1008 and the AC coefficient quantized by the quantization circuit 1005, and outputs it as coded data.

The storage circuit 1010 writes the code data (that is, the compressed image data) output from the entropy coding circuit 1009 into a memory or a hard disk and stores it. Further, at the time of outputting an image, the storage circuit 1010 reads the stored code data and outputs it to the entropy decoding circuit 1011.

The entropy decoding circuit 1011 decodes the coded data output from the storage circuit 1010 to restore the combined DC coefficient and AC coefficient. The area tag signal separation circuit 1012 separates the combined DC coefficient into the area tag signal and the DC coefficient which is the original image data. The separated area tag signal is supplied to the image output device 1018. The separation process will be described later.

Next, the dequantization circuit 1015 dequantizes the DC coefficient output from the area tag signal separation circuit 1012 and the AC coefficient output from the entropy decoding circuit 1011 based on the dequantization table. . The inverse DCT calculation circuit 1016 performs inverse DCT conversion on the DC coefficient and the AC coefficient which are inversely quantized by the inverse quantization circuit 1015. As a result, the frequency space data is converted into blocked image data. Block / raster circuit 101
Reference numeral 7 converts the blocked image data into raster data.

The image output device 1018 responds to the area tag signal output from the area tag signal separation circuit 1012.
Image processing suitable for the attribute of the block is performed on the image data, and a hard copy is output.

(2) Area Tag Signal Next, the area tag signal of this embodiment will be described. Original images input from the image input apparatus 1001 often include images having different properties. For example,
Areas such as letters and line drawings that contain a lot of high-frequency components, areas such as business graphs that have a relatively small number of colors and have a geometrical structure, or areas such as natural images / photographs that are high in color. Although there are few frequency components, there are areas where pixel values change in a complicated manner on a pixel-by-pixel basis. In addition, there may be a background area in which nothing is drawn. Furthermore, DT
An original image created by an application program such as P (desk top publishing) has a region in which the pixel level constantly changes, such as a computer-generated region that does not include noise at all or a photo region read by a scanner. Are mixed.

Generally, when a high quality image output is required in a digital copying machine or the like, image quality correction such as color correction and gradation correction is added to the image data according to the performance of the device. Further, the image size may be scaled as necessary. As described above, when areas having greatly different attributes are mixed in one original image, in order to output such an original with high quality and high image quality, an image quality correction suitable for each area is performed. Is desirable. The area tag signal is a signal generated for identifying areas having different attributes and performing different processing, and is generated for each block. Also,
There are four types of attributes of the image area identified in the present embodiment: character / line drawing, graphic, photograph / natural image, and background, and these can be identified based on variations in pixel values in the target block. . Since these four types of attributes can be represented by 2 bits, the area tag signal is 2
Consists of bits.

(3) Region Tag Signal Combining Process Next, the combining process of the region tag signal and the DC coefficient according to the present embodiment will be described with reference to FIG. In the ADCT encoding, generally, the DC coefficient and the AC coefficient, which are the coefficients of the frequency space obtained by the DCT transform, are quantized by the quantization table, and then the entropy encoding is performed. In the JPEG baseline system, which is the standard encoding system for color facsimiles, the output data width of the DCT arithmetic circuit is 11 bits. Huffman coding is used for entropy coding, and the input data width is 11 bits.

Further, in the JPEG baseline system, 16 is recommended as the quantization value of the DC coefficient as the value of the quantization table. That is, D obtained as a result of the DCT operation
By dividing the value of the C coefficient by 16, the resulting 11-bit data is quantized into 7-bit data. Since the input data width of the entropy coding circuit 1009 for inputting the quantized DC coefficient is 11 bits, there is a margin of 11 bits-7 bits = 4 bits in the upper bit portion. Area tag signal synthesis circuit 1008
Embeds a 2-bit area tag signal in the bit width portion in which the margin is generated to synthesize the area tag signal and the DC coefficient.

In this case, the D in which the area tag signal is embedded
Since the C coefficient fits within the input bits of the entropy coding circuit 1009 (that is, 7 bits + 2 bits <1).
1 bit), the existing Huffman coding circuit based on the JPEG baseline system can be used as it is without adding a special function to the entropy coding circuit 1009. Since the quantized value of the DC coefficient recommended by the JPEG baseline method is 16, it is possible to hold a region tag signal of up to 4 bits when using the recommended quantized value.

Here, the DC coefficient indicates the average density of the block, and the DC components of adjacent blocks are correlated with each other. From this, in the JPEG baseline method, the difference between the DC component of the target block to be encoded and the DC component of the immediately preceding block is calculated, and this difference value is entropy encoded to obtain the DC component. Is encoded. Therefore, when the target block and the previous block have the same area tag signal, the area tag signal is canceled by the calculation of the difference.
As a result, even if the area tag signal is added to the DC component, the overall code amount does not increase as much as the addition of the area tag signal, so that the influence on the compression rate is reduced. Also, JPE
Since the G-baseline type DCT operation circuit, the quantization circuit, the Huffman coding circuit and the like can be used as they are, they can be realized only by adding the area tag signal synthesis circuit 1008 to the existing configuration.

(4) Region Tag Signal Separation Process Next, the region tag signal separation process according to the present embodiment will be described with reference to FIG. Entropy decoding circuit 1
011 adds the decoded DC coefficient and the already decoded DC coefficient of the previous block, and the result is D
Output as C coefficient. Area tag signal separation circuit 1012
Is the D output from the entropy decoding circuit 1011.
The upper 2 bits of the C coefficient are separated and output as the area tag signal, and the remaining 7 bits are output as the DC coefficient of the original image data.

This DC coefficient and entropy decoding circuit 1
Inverse quantization is performed by the inverse quantization circuit 1015 together with the AC coefficient output from 011 and image data is restored by performing inverse DCT operation by the inverse DCT operation circuit 1016, and the result is passed through the block / raster circuit 1017. And outputs it to the image output device 1018.

The image output device 1018 receives the area tag signal output from the area tag signal separation circuit 1012 and the image data output from the block / raster circuit 1017. The area tag signal consists of 2 bits as described above, which allows the image output apparatus 1018 to discriminate four types of attributes and perform a correction process suitable for each attribute. For example, for the character / line drawing area, high-quality character / line drawing output can be reproduced by emphasizing the edge of the image and removing the background in the background. Also,
In the photograph / natural image area, unlike the character / line drawing area, a high quality image can be reproduced by smoothing the image. Furthermore, with respect to the background area, a flat background can be reproduced by largely removing the background. In this way, it becomes possible to output a high-quality / high-quality image that cannot be obtained when a single image correction is applied to the entire image.

B: Second Embodiment (1) Overall Structure Next, a second embodiment of the present invention will be described. FIG.
FIG. 6 is a block diagram showing an overall configuration of an image processing apparatus which is a second embodiment of the present invention. Unlike the first embodiment described above, the present embodiment is configured on the assumption that a full-color original image is input. In FIG. 4, an image input device 1301 is, for example, a scanner, which reads an image of a document and converts it into RGB color image data (digital data). The image input device 1301 may be an external interface or the like that takes in RGB image data from an external network.

The color space conversion circuit 1302 converts the color space of the input image data from RGB to YCbCr of the luminance color difference system. In addition, the raster / block circuit 1303 uses YCb
The input image data converted into Cr is divided into pixel blocks each having 8 × 8 pixels as a unit.

Next, the area tag signal generation circuit 1304
The attribute of the image data is determined for each block, and the determination result is output as the area tag signal of the block. DCT
The arithmetic circuit 1305 converts the image data into frequency space data in block units, and outputs DC coefficients and AC coefficients.

Next, the quantization circuit 1306 quantizes the DC coefficient and the AC coefficient based on the values of the quantization table stored in the rewritable memory. Here, the quantization circuit 1306 corresponds to each attribute of character / line drawing, photograph / natural image, graphic, and background, for example, for a luminance signal (FIG. 12A) and color difference as shown in FIG. It holds a quantization table for signals ((b) in the figure). In addition, the quantization table switching circuit 1307 selects a combination of the quantization table for the luminance signal and the color difference signal corresponding to the attribute of the input image based on the area tag signal supplied from the area tag signal generation circuit 1304, The quantization table is switched according to either the luminance signal or the color difference signal. The code amount of the quantized data of the color difference signal may be reduced by sub-sampling in half in the main scanning direction.

Next, the area tag signal synthesizing circuit 1310
The area tag signal supplied from the area tag signal generation circuit 1304 and the DC coefficient of the luminance signal are combined. The synthetic process will be described later. Further, the entropy coding circuit 1311 Huffman-codes the combined DC coefficient output from the area tag signal combining circuit 1310 and the AC coefficient quantized by the quantization circuit 1306, and outputs the obtained code data. Further, the storage circuit 1312 writes and stores the code data output from the entropy coding circuit 1311 in a memory or a hard disk.

Next, the entropy decoding circuit 1313
Decodes the code data read from the storage circuit 1312 at the time of image output, and restores the composite DC coefficient and AC coefficient. The area tag signal separation circuit 1314 separates the combined DC coefficient into the area tag signal and the DC coefficient of the luminance signal.
The separated region tag signal is quantized by the quantization table switching circuit 1
317 and the image quality correction circuit 1321.

Next, the inverse quantization circuit 1318 inversely quantizes the DC coefficient and the AC coefficient based on the values in the quantization table. The quantization table switching circuit 1317 selects a quantization table set based on the region tag signal and switches the quantization table according to either a luminance signal or a color difference signal, as in the case of encoding. Details of switching the quantization table will be described later. When the color difference signal is sub-sampled at the time of encoding, the inversely quantized color difference signal is expanded in the main scanning direction, and both the luminance signal and the color difference signal are converted into 8 × 8 block size data.

Next, the inverse DCT calculation circuit 1319 converts the frequency space data of the luminance signal and the color difference signal into the original image data. In addition, the block / raster conversion circuit 132
0 converts a block of image data into raster data.

Next, the image quality correction circuit 1321 switches the content of the correction processing according to the area tag signal supplied from the area tag signal separation circuit 1314, and performs the image quality correction processing suitable for the image block of each attribute. In addition, the image quality correction circuit 13
Reference numeral 21 denotes the image output device 1 that receives the image signal received by the YCbCr.
It is converted into a YMCK signal that matches the characteristics of 322. Then, the image output device 1322 performs screen processing suitable for the image attribute on the image data converted into the YMCK signal, and outputs this.

(2) Switching of Quantization Tables Next, switching of quantization tables according to this embodiment will be described. When quantizing the DC coefficient and the AC coefficient obtained as a result of the DCT operation in ADCT encoding,
The value of the quantization table is quantized as a quantization step, and the value of this quantization table affects the image quality of the decoded image. That is, the smaller the quantization step is, the less the image quality is deteriorated. However, for example, when “1” is set as the quantization value, it is substantially the same as no quantization. Therefore, in this case, the code amount and the entropy coding circuit 1313 are used. Is not practical because it does not reduce the load on the On the other hand, depending on the attribute of the image, there are some that are not significantly affected by this quantization, but some are greatly affected. For example, in an image area having an attribute such as a photograph / natural image, since high frequency components are not included so much, increasing the quantized value of the high frequency components has little effect on the image quality. Since many frequency components are included, it is desirable to reduce the quantization value of the high frequency components in order to obtain a high quality decoded image.

Therefore, in the present embodiment, for a photograph / natural image, for example, a quantization table as shown in FIG. 5A, and for a character / line drawing, for example, a quantization table as shown in FIG. 5B. A quantization table suitable for the image attribute such as a table is prepared, and a high-quality decoded image is obtained by switching these. That is, the quantization table switching circuit 1307 selects an optimum quantization table from a plurality of prepared quantization tables based on the area tag signal generated by the area tag signal generation circuit 1304 and uses it. To perform quantization processing. As described above, according to the present embodiment, it is possible to switch the quantization processing corresponding to the image attribute with a relatively simple method. The switching of the quantization processing as in the present embodiment can be applied to the case of the monochrome original as in the first embodiment described above.

(3) Region Tag Signal Combining / Separating Process Next, the region tag signal combining / separating process according to the present embodiment will be described. When encoding full-color image data as in this embodiment, image data of three planes of YCbCr signals are treated as one set. Therefore,
If the area tag signal is combined with the DC component of the luminance signal, there is no need to combine the area tag signal with the DC coefficient of the color difference signal, and therefore the area tag signal combining circuit 1 of the present embodiment.
310 combines the area tag signal only with the DC coefficient of the luminance signal. On the other hand, the DC coefficient of the color difference signal is Huffman-encoded without being combined with the area tag signal.

The area tag signal may be combined with the DC coefficient of the color difference signal instead of being combined with the DC coefficient of the luminance signal as in the present embodiment. Further, the area tag signal synthesis circuit 1310 and the area tag signal separation circuit 1
For the synthesis / separation in 314, the same method as in the first embodiment can be adopted.

C: Third Embodiment Next, a third embodiment of the present invention will be described. FIG.
FIG. 9 is a block diagram showing an overall configuration of an image processing apparatus which is a third embodiment of the present invention. It should be noted that the present embodiment is configured on the assumption that a monochrome original image is input. In the present embodiment, the unit of the pixel block for discriminating the attribute of the image area is set to 4 × 4 pixels, and therefore an improvement is made to the resolution of the attribute discrimination in comparison with the first embodiment shown in FIG. It is a thing. Specifically, a block division circuit 1501 is added to the latter stage of the raster / block conversion circuit 1303, and the area tag signal generation circuit 1003 and the area tag signal synthesis circuit 1 are added in order to cope with the resolution improvement.
Instead of the area tag signal separation circuit 1012 and the image output apparatus 1018 (see FIG. 1), the area tag signal generation circuit 1503, the area tag signal synthesis circuit 1508, the area tag signal separation circuit 1512, and the image output apparatus 1518.
(See FIG. 6). Other configurations are similar to those in FIG.

The operation of this embodiment will be described below. First, in the first embodiment, the attribute of an image is discriminated in block units of 8 × 8 pixels, but in the present embodiment, as shown in FIG. 7, an 8 × 8 pixel block is divided into four, and 4 × 4 pixels are divided. The determination is made in sub-block units. In the present embodiment,
There are two types of attributes of the image area to be discriminated, that is, character / line drawing area and photograph nature / image area, and the discrimination result is represented by 1 bit. This is because the margin of the input data width of the entropy coding circuit 1009 is 11 bits−7 bits = 4 bits, and therefore, to hold all the determination results for four sub blocks, 1 is set for one sub block. This is because it needs to be expressed in bits.

Now, as shown in FIG. 6, the 8 × 8 pixel block output from the raster / block conversion circuit 1002 is divided into four sub blocks composed of 4 × 4 pixels in the block division circuit 1501. To be done. And
In the area tag signal generation circuit 1503, the image attribute of each sub block is determined based on the variation of the pixel value in each sub block, and a 1-bit area tag signal is generated for each sub block. The area tag signals for each sub-block are collected as a 4-bit area tag signal, and the area tag signal synthesis circuit 1508 outputs 8 × 8.
It is combined with the DC coefficient of the pixel block. The other encoding process is the same as in the first embodiment.

On the other hand, at the time of decoding, the 4-bit area tag signal is DC by the area tag signal separation circuit 1512.
It is separated from the coefficient and supplied to the image output device 1518. Then, the image output device 1518 discriminates the attribute of the image in units of 4 × 4 pixels based on each 1 bit forming the 4-bit area tag signal, and performs a correction process suitable for the attribute on the corresponding sub-block. The other decoding process is the same as that of the first embodiment.

As described above, according to this embodiment, a higher quality output image can be obtained as compared with the case where the attribute is discriminated in the unit of 8 × 8 pixels as in the first embodiment.
For example, in the case where an attribute determination is made on a region in which a character region and a photo region are mixed in an original image in units of 8 × 8 pixel blocks, if the block includes a boundary between the character region and the photo region,
In the first embodiment, there is no choice but to discriminate as one of the attributes. On the other hand, in the present embodiment, 8 × 8
By further dividing the pixel block into sub-blocks of 4 × 4 pixels, the attribute near the boundary can be finely discriminated, the discrimination accuracy is improved, and as a result, a high quality image output can be obtained. Further, even when the output image is enlarged, the attribute discrimination is highly accurate, so that the image quality deterioration near the boundary due to the enlargement processing can be reduced. In the present embodiment, a case where a monochrome original is input is taken as an example for simplicity, but a full-color original can be processed in the same manner.

[0058]

As described above, according to the present invention, an input image in which regions having different attributes such as characters / line drawings, photographs / natural images and the like are mixed is separated, ADCT-encoded and accumulated, and then decoded. In the case of converting and outputting, the identifier (region tag signal) indicating the attribute of the image region is combined with the DC coefficient of the pixel block, entropy-encoded, and the identifier is separated after decoding. No additional means needed. In addition, DC at the time of encoding
Since the information of the same identifier is canceled by the coefficient difference calculation, it is possible to suppress the increase in the code amount due to the addition of the identifier as much as possible. As a result, it is possible to suppress a decrease in compression efficiency associated with the addition of an identifier, avoid an increase in circuit scale and an increase in cost, enable highly accurate area separation, and perform image processing suitable for each image area. It is possible to obtain a high-quality output image (claims 1 to 7).

Further, according to the invention described in claim 5, since the image data is quantized by the quantization step suitable for the attribute of the image area, it is possible to suppress the image quality deterioration due to the quantization.
Further, according to the invention described in claim 7, since the resolution of the attribute discrimination is improved, a higher quality output image can be obtained.

[Brief description of drawings]

FIG. 1 is a block diagram showing a configuration of a first embodiment of the present invention.

FIG. 2 is a diagram for explaining a process of combining region tag signals in the same embodiment.

FIG. 3 is a diagram for explaining a separation process of a combined tag signal in the same embodiment.

FIG. 4 is a block diagram showing a configuration of a second embodiment of the present invention.

FIG. 5 is a diagram illustrating a quantization table used in the embodiment, (a) shows a quantization table for a photograph / natural image, and (b) shows a quantization table for a character / line drawing. Is shown.

FIG. 6 is a block diagram showing a configuration of a third embodiment of the present invention.

FIG. 7 is a diagram for explaining a process of generating a region tag signal in the same embodiment.

FIG. 8 is a block diagram showing a process of ADCT encoding.

FIG. 9 is a diagram showing an example of an 8 × 8 pixel block.

10 is a diagram showing DCT coefficients obtained by DCT transforming the 8 × 8 pixel block shown in FIG. 9.

11 is a diagram showing a value obtained by quantizing the DCT coefficient shown in FIG.

FIG. 12 is a diagram showing an example of a recommended quantization table in the JPEG baseline method, where (a) shows a recommended quantization table for luminance signals and (b) shows a recommended quantization table for color difference signals. Shows.

FIG. 13 is a diagram showing the order of zigzag scanning.

FIG. 14 is a block diagram showing a first configuration example of a conventional device.

FIG. 15 is a block diagram showing a second configuration example of a conventional device.

FIG. 16 is a block diagram showing a third configuration example of a conventional device.

[Explanation of symbols]

 1001 image input device 1002 raster / block circuit 1003 region tag signal generation circuit 1004 DCT operation circuit 1005 quantization circuit 1008 region tag signal synthesis circuit 1009 entropy encoding circuit 1010 storage device 1011 entropy decoding circuit 1012 region tag signal separation circuit 1015 reverse Quantization circuit 1016 Inverse DCT operation circuit 1017 Block / raster circuit 1018 Image output device

Claims (7)

[Claims] 1. An image encoding apparatus for ADCT encoding image data, a block forming unit for dividing input image data into blocks, and an identifying unit for identifying a feature of each block and generating an identifier indicating the feature. A DCT transforming unit that performs DCT transform on each block to generate a DC coefficient and an AC coefficient; a quantizing unit that quantizes the DC coefficient and the AC coefficient; and the quantized DC coefficient and the identifier. And combining means for generating a combined coefficient, and encoding means for entropy-encoding the difference value between the quantized AC coefficient and an adjacent block of the combined coefficient to generate coded data. A characteristic image encoding device. 2. An image decoding apparatus for decoding coded data generated by the image coding apparatus according to claim 1, wherein the coded data is decoded and the quantized AC coefficient and the synthesizing are performed. Decoding means for restoring coefficients, separation means for separating the restored combined coefficient into the quantized DC coefficient and the identifier, and inverse quantum for inverse quantizing the quantized AC coefficient and DC coefficient. And DCT for converting the inversely quantized DC coefficient and AC coefficient.
An image decoding apparatus comprising: an inverse DCT conversion unit that converts and restores image data of each block. 3. The image coding apparatus according to claim 1, storage means for storing code data coded by the image coding apparatus, and decoding code data read from the storage means. 2. An image processing device comprising: the image decoding device according to 2. 4. The synthesizing unit generates a synthesizing coefficient by adding the bit string of the identifier to the bit string of the DC coefficient, and the separating unit generates the bit string of the synthesizing coefficient decoded by the decoding unit. 4. The bit string of the DC coefficient and the bit string of the identifier are separated.
An image processing apparatus according to claim 1. 5. The quantizing means selects and quantizes a quantizing step corresponding to the feature of each block based on the identifier generated by the identifying means, and the dequantizing means uses the dequantizing means. The image processing apparatus according to claim 3, wherein an inverse quantization step corresponding to the quantization step by the quantization means is selected and inversely quantized based on the identifier decoded by. 6. The image processing apparatus according to claim 3, further comprising: an identifier decoded by the decoding unit for the image data of each block restored by the inverse DCT conversion unit. An image processing apparatus comprising: a correction unit that performs a correction process corresponding to. 7. The identifying means further divides the block into sub-blocks, identifies features for each sub-block, and generates respective identifiers, and the correcting means subdivides the image data of each block into sub-blocks. The image processing apparatus according to claim 6, wherein correction processing corresponding to the characteristic indicated by the identifier is performed in block units.