JP3227181B2 - Image processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to compression and decompression of color image data.

[0002]

2. Description of the Related Art Conventionally, an image is converted into a DC by 8 × 8 block units.
A JPEG standard coding method is known in which the data is quantized and Huffman-coded using a single quantization table. On the other hand, a unique coding method for performing quantization and Huffman coding by switching a quantization table for each 8 × 8 block according to the characteristics of an image is also known. According to this method, it is possible to perform efficient encoding in consideration of the characteristics of the image, as compared with the standard encoding method.

[0003]

If suppression of image quality deterioration / encoding efficiency is taken into consideration, it is preferable to use the above-mentioned unique encoding system as much as possible. However, the use of the unique encoding method has a problem that compatibility with general devices that only support the standard encoding method cannot be obtained.

[0004] The present invention has been made in view of the above-mentioned problems, and efficiently performs efficient encoding in which quantization is performed using a quantization table corresponding to a feature for each block of an image.
Further, it is another object of the present invention to be able to cope with an encoding method that performs quantization regardless of the characteristics of each block. In particular, the coded data obtained by the former coding method can be transmitted immediately, and the coded data obtained by the latter coding method can be transmitted by generating from the coded data obtained by the former coding method. One purpose is to do. Furthermore, the present invention focuses on the former encoding method, which is a method of changing a quantization table for each block.In the former encoding, while performing quantization to some extent according to the characteristics of an image, It is also an object to make the amount of encoded data obtained here a desired amount.

[0005]

According to the present invention, there is provided an image processing apparatus comprising: an input unit for inputting image data; a dividing unit for dividing the image data into a plurality of blocks; A frequency conversion coefficient for each block, a determination means for determining the characteristics of the image of each block, and a frequency conversion coefficient for each block obtained by the conversion means. Encoding means for selectively using the quantization and encoding to obtain the first encoded data, and encoding data encoded and encoded by the transmission partner based on the selective use of the quantization table for each block. The first coded data and the data indicating the determination result are transmitted or the first coded data is decoded and a single quantization table Control means for transmitting or controlling the second encoded data obtained by re-encoding using the first encoded data, wherein the determining means sets the code amount of the first encoded data to a predetermined code amount Therefore, the determination result is
It is characterized in that correction is made according to the result of comparing the input amount of image data by the input means and the generated code amount by the encoding means.

[0006]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention relates to a color image compression / decompression. In particular, JPEG (Joint Photograph)
ic Expert Group) A method for solving the problems when the standard coding method is used, and a new method for connecting a device having a constant-speed recording with good consistency, such as an electrophotographic recording device. Is provided.

In the JPEG coding method, the quantization table and Huffman table in one page are fixed, so that a photographic image having a low frequency component or a character image having a high frequency component is mixed in the original. However, the restored image was not always a good image. In color facsimile and color DTP communication, it is natural that there are many mixed images in a document, and even if a standardized encoding method is used unless such an image is properly restored, it is difficult to use in this field. However, in the field of exchanging information by communicating with the other party, such as color image coding, using a unique coding method impairs communication over a wide area, and the general idea is to adhere to the standard method as much as possible. SUMMARY OF THE INVENTION The present invention provides a new compression communication system which is capable of transmitting high image quality and high coding efficiency in the in-house mode, and has a system capable of converting from the in-house mode for communication using the JPEG standard compression system.

[0008] Based on the above concept, the present invention uses a standard encoding system for communication between different types of devices. If a high definition image is to be transmitted, the encoding efficiency is reduced. Transmit with reduced image quality. If the other party is the same model, high quality and high coding efficiency transmission can be performed in the company mode.

In the embodiment of the present invention, the encoding table is appropriately changed in units of 8 × 8 blocks according to the image area determination amount indicating the type of image, and the image area information and the compressed data are stored. If the communication partner requests compressed data in the standard mode, 1
An optimum encoding table is set based on the image area information for the page. Alternatively, an encoding table specified by the user from the operation panel is used. Once the encoding table is set,
Decompress using image area information and compressed data. The decompressed restored data is encoded in the standard mode using the above-described encoding table without changing the table for one page. The encoded data is stored in another compressed page memory. The compressed data stored for one page is transmitted to the other side through the transmission path.

On the other hand, if an actual memory of one page is provided for interfacing with a device that requires a relatively high recording speed, such as electrophotographic recording, and requires a constant recording speed, the cost is high and practical use is not possible. Therefore, a method is used in which recording data is transmitted to a recording device while being stored in a compression memory and decompressed. At this time, a problem arises in that the amount of information of image data transmitted from the host differs depending on the original and the image quality, so that a fixed compression memory overflows in some cases and leaves extra memory in some cases. Therefore, it is required to control the amount of memory constant. In this case as well, a method is provided that takes into account both the company's own mode and the standard mode, and enables mutual conversion. That is, the compressed image stored in the host has reduced coding efficiency, improved image quality, and sometimes has a large data amount.
In some cases, the code efficiency is improved and the data amount is small. Further, the data amount of the same image differs depending on the compression method.
However, in a recording method such as electrophotographic recording that requires constant-speed recording at the time of recording, image data for recording is usually recorded in a fixed semiconductor memory and then transferred to an electrophotographic recording device. Therefore, if the compressed data from the host constantly changes, buffering for electrophotographic recording cannot be performed. The present invention enables image data from an input scanner and compressed or uncompressed data from a host to be recorded in a high-quality and fixed buffering memory, and can be connected to a constant-speed recording device with a small amount of memory and inexpensively. An apparatus is provided.

The method for realizing the present invention comprises the following steps: image position information of an original image (for example, the advance of an address), advance of compressed data (also the advance of an address), and image area information of an 8 × 8 block to be compressed. An encoding table for compressing the inside of the block is selected from the three. If the compressed data advances more slowly than the original image data, and if the next block to be compressed is a gradation image with few high-frequency components, the next encoding table is used as the next encoding table, and the table that greatly enhances the compression efficiency is used. select. If the block contains a character image or the like having a high frequency component, an encoding table that slightly increases the compression efficiency is selected. Therefore, the optimal encoding table determined from the image area in the block also differs from the table actually encoded by the memory control. Therefore, the image area information memory is modified by the memory amount control circuit. In this way, the amount of compressed memory is kept constant regardless of the information amount of the original image. The above is the outline of the constant memory control method.

Further, the method for realizing the present invention comprises the following steps: the image position information (for example, the advance of the address) of the original image from the input device or the decompressed image from the host and the advance of the compressed data or recompressed data (also the address) An encoding table for compressing the inside of the block is selected from three of the image area information of the 8 × 8 block to be compressed. The memory usage of the compressed data is faster than the progress of the processing of the pre-compression image data, that is, when the compression efficiency is inferior to a certain amount of memory, and the next block to be compressed is a gradation image with a small high frequency component. If there is, the next encoding table is selected, and a compression table that greatly enhances compression efficiency is selected. If the block contains a character image or the like having a high high-frequency component, a compression table that slightly increases the compression efficiency is selected. Therefore, the actual compression table is determined based on the intra-block image area information and the use speed of the compression memory for the progress of the input image. In this way, the amount of compressed memory is kept constant regardless of the amount of information of the original image or the amount of image information from the host. The above is the outline of the constant memory control method of the present invention.

Next, an embodiment of the present invention will be described in detail with reference to the drawings.

FIG. 1 is a block diagram showing the overall configuration of the image processing apparatus according to the present embodiment. In FIG. 1, reference numeral 1 denotes a color copying machine, which includes a color scanner 101 and a color printer 102. In addition, it has a controller for controlling both. Reference numeral 2 denotes a color copier interface, which is a color workstation 119 described later.
And the color copier 1, and further performs processing such as trimming, compression, storage, decompression, and scaling of image data inside. Reference numerals 103 and 122 denote video interfaces for connecting the color copying machine 1 and the color copying machine interface 2, and communicate image data and command data, respectively. Reference numeral 104 denotes a trimming circuit which outputs a designated portion of the input image to 105 and thereafter. Similarly, when outputting image data, the designated portion is
For recording an image. A conversion circuit 105 converts the RGB data input from the color scanner 101 into rgb data in a standard color space.
It is composed of a table converter such as AM and ROM. 106
Conversely, a conversion circuit converts rgb data in a standard color space into YMCK signals matched to the color characteristics unique to the printer. This is also usually converted by a table converter. 10
A selector 7 selects the direction of input / output data. 108
Is a double buffer memory, which is basically composed of a line memory for a JPEG standard 8-line X pixel scanning width.
For example, if 16 pels A4 3 colors and width are 8 × 16
× 210 × 2 × 3 = 161.3 Kbytes of memory is required. A circuit 109 frequency-converts image data by DCT (discrete cosine transform) for each 8 × 8 block. Reference numeral 132 denotes an image area determination circuit that determines the type (character, photograph, halftone dot, etc.) of the original image for each 8 × 8 pixel block.

The image area determination circuit 132 can determine the image area by various methods such as a method of determining from the relationship between the DC portion and the AC portion of the DCT circuit 109. This image area determination amount gives a large value in an image region in which the target portion includes many high-frequency components such as character thin lines, and in an image in which the frequency components are many in low-frequency components such as an image region of a human face portion. Give a small value. As the image area determination amount, for example, 3-bit data is given. The memory 112 determined in this way is 8
Stores image area determination data of × 8 blocks. The memory amount of this part is 2 × 210 × 16 × 8 / (8 × 8 × 2) =
420 bytes of memory is required. 110 is 1
This is a compression circuit that selects an optimal quantization table and Huffman table according to the image area determination amount of 32, and compresses the data by the JPEG ADCT method. That is, since a table that is in the image area of the current block is selected and coded for each 8 × 8 block, the quantization table is no longer fixed standard coding, but to improve image quality and coding efficiency. It is extremely effective. The compressed data and the image area information data in the memory 109 are stored in a compressed page memory 111 for one page and an image area information table 112 for one page. As will be described later, if the transmission destination is a host computer capable of decoding the original compression method for switching the quantization table, the compressed data and the image area information data are transmitted to the host through the scsi interface 116. On the other hand, if the transmission destination is a host that cannot decode the original compression method, the data is converted to the standard JPEG method and transmitted as follows.

Reference numeral 113 denotes a compressed data decompression circuit. Since the quantization table used for compression is determined by using the image area information table 112 at the time of decompression, the decompression is performed by using the inverse quantization table corresponding to the table at the time of decompression.

The decompressed and restored image data is written into the double buffer memory 108 while being reduced and enlarged through a scaling circuit. The written restored image is entirely compressed on one page by using a quantization table designated manually or automatically. This method is a JPEG standard method.
The compressed data is transmitted to the host through the SCSI of 116 through the compression buffer memory of 115. At this time, the image area information data 112 is not transmitted to the host side, but the compression table data for one page is transmitted. Reference numeral 119 denotes a host computer which constitutes a color workstation, and operates under a Unix, X-window environment. The language uses, for example, the C language, 120 is a SCSI interface, and 121 is coding software that operates in software. Reference numeral 124 denotes an application program for controlling the color copying machine 1 and the color copying machine interface 2. Thus, the color image from the color scanner 101 can be compressed and taken into the host computer 119.

Reference numeral 117 denotes a controller for controlling each section of the color copier interface, and 118 denotes a command interpreter for analyzing commands from the color workstation and generating control signals. In addition, for example, an error is notified to the host computer, and an operation of interlocking the whole is performed. 129, 130, 13
1 is a counter for measuring the data amount, 132 is a selector for selecting one of the counters 129 and 131, and 115 is data for controlling the data amount according to the count values of the counters 129 to 131. Quantity controller. Reference numeral 133 denotes an operation unit for manually selecting an encoding parameter (a quantization table, a Huffman table, and the like).

The processing performed by the above configuration will be described.

(1) Image Input from Color Scanner 101 The color scanner 101 is composed of R (red), G (green), and B (blue) line sensors, scans an original, and outputs an 8-bit digital signal. Occurs. The generated digital signal is input to the color copier interface 2 via the interface 103 in RGB dot sequence for each pixel.

(2) Internal Processing of the Color Copier Interface 2 The RGB signal for each pixel input from the color scanner 101 is trimmed only in a necessary area by the trimming circuit 104, and r and g in the standard color space are converted by the conversion circuit 105. , B signals, and at the same time, the r, g, b signals are sent to the double buffer 108 as parallel signals. In the subsequent processing, similar processing is performed in parallel on each of the r, g, and b signals.

The counter 129 measures the amount of input image data.

(3) Data Compression Data compression in this embodiment is basically performed as follows.

Variable length compression is used as an algorithm for data compression. For example, JPEG (Joint
Photographic Expert Grou
p) ADCT (Adaptive Discrete)
As in the case of the Cosine Transform method, image data is subjected to frequency conversion for each block, quantized by a predetermined quantization parameter, and then a DC component in the block is subjected to Huffman coding of a prediction error, and an AC component is zigzag. Scan and perform Huffman coding.

FIG. 2 shows the configuration of the compression circuit 110.

In FIG. 2, reference numeral 201 denotes a quantizer for performing linear quantization on an inputted frequency transform coefficient, 202 denotes a ROM for storing the quantized coefficient, and 203 denotes Huffman coding of the quantized transform coefficient. Huffman encoding unit.

The quantized coefficients stored in the ROM 202 are 8 × 8 two-dimensional matrices of eight types M ₁ to M ₈ . Nears from M ₁ to M _8, the value of the quantized coefficients of the high frequency components becomes larger, it is predetermined as the quantization step becomes larger. The eight types of matrices are selected based on the image area determination data (3 bits) from the image area determination circuit 132. By selecting the quantization matrix in accordance with the characteristics of the image to be compressed in this way, it is possible to suppress the deterioration of the image quality and perform data compression at a high compression rate.

The above data compression is performed by the color scanner 10.
1 can be performed in parallel with the image reading. The compressed image data is stored in the compression memory 111. At this time, the counter 130 measures the amount of data stored in the memory (the free space of the memory) by counting the addresses of the compression memory.

On the other hand, the image area determination data for each block is stored in the memory 112.

(4) Data Decompression Data decompression is performed on the compressed image data stored in the compression memory 111 by using the image area determination data for each block in a procedure reverse to that of data compression.

At this time, the counter 130 measures the amount of data stored in the memory (the free space of the memory) by counting the addresses of the compression memory. On the other hand, the counter 131 measures the amount of decompressed image data.

(5) Magnification The expanded image data is supplied to a magnification circuit 114 as necessary.
Is subjected to a scaling process and sent to the double buffer 108. After the double buffer 108, a loop is formed, and the same processing can be performed a plurality of times.

By providing this scaling circuit 114,
For example, it is possible to obtain a magnification that exceeds the magnification that is possible when reading an image with the color scanner 101.

(6) Image Area Judgment The characteristics of the input image (the degree of being a character or a photograph) are judged by the image area judgment circuit 132.

The determination is basically performed by dividing the magnitude of the high-frequency component of the transform coefficient frequency-converted for each block in the DCT section 109 into eight stages for each block.
In this manner, three bits of image area determination data indicating the degree of a character image having many high frequency components are generated.

The image area determination circuit 132 has a configuration as shown in FIG. Reference numeral 301 denotes a high frequency component detection circuit,
By examining the AC part of the conversion coefficient, image area determination data indicating the magnitude of the high frequency component is generated. Reference numeral 302 denotes a determination data correction circuit that corrects image area determination data according to a correction signal from the data controller 115.

When image data is input from the color scanner 101, the data controller 115 monitors the count values of the counters 129 and 130, and corrects by comparing the input amount of image data with the increase amount of compressed data. Generate a signal.

For example, when the increase amount of the compressed data is larger than the input amount of the image data, the judgment data is increased by 1 in order to increase the compression ratio. That is, when should use a table of M ₅ is modified to use the table of M _6. The correction signal can be a multi-level signal according to the relationship between the input amount and the increase amount.

Similarly, when the amount of compressed data stored in the compression memory 111 is changed and stored again in the compression memory 111, a correction signal is generated by comparing the count values of the counter 130 and the counter 131.

The data amount controller 115 generates a correction signal according to the size information of the document input from the command interface 122 and sent via the controller 117. In this embodiment, the color scanner 101 reads A3 and the compression memory 111 also reads A3.
It has the capacity of one sheet. Here, for example, when an A4 document is read, the size information detected by the document size detection is sent to the data amount controller, and control is performed in consideration of the size information.

More specifically, since the amount of color image data input from, for example, a scanner or a host computer is specified in advance, the total amount of image data to be input can be calculated in advance. On the other hand, the amount of compressed image data memory is also fixed at a fixed value because it is fixed in hardware. For example, if the input image data amount is 10 Mbytes and the compression memory amount is 1 Mbytes, the input data advances to 5 Mbytes and the compressed data becomes 0.6 Mbytes.
If it is bytes, the compression efficiency is poor, and thereafter, a command is issued from 115 to increase the compression efficiency. Therefore, even if the image area determination circuit in 109 determines that the image is a high-frequency component image, a compression table is selected that increases the compression efficiency even when the image quality is lower than when the instruction from 115 does not require high compression efficiency. Conversely, even if the input data reaches 0.5 Mbytes, if the output data is 0.4 Mbytes,
For example, a compression table is selected that increases the image quality and lowers the compression efficiency even when an image with a small high-frequency component comes. In this way, the amount of compressed memory is controlled to be constant regardless of the amount of information of the original image.

(7) Input from Computer The host computer 119 sends a SCSI I / F1
16 allows input of raw image data and compressed image data.

When the input is raw image data,
The data is sent from the SCSI I / F 116 to the double buffer 108, and the subsequent processing is as described above.

On the other hand, when the compressed image data is input, whether the compression changes the quantization parameter according to the image area as described above (hereinafter referred to as “independent mode”), Is processed as follows depending on whether is a fixed one (hereinafter, referred to as “standard mode”).

In the case of original mode compressed data, the SCS
The compressed image data is stored in the compression memory 111 through the I / F 116, and the image area determination data for each block is stored in the memory 112.

On the other hand, in the case of standard mode compressed data,
The compressed image data is stored in the compression memory 111, and a quantization table (one type) for one screen is stored in the memory 112. In this case, the image data can be decompressed once by the decompression circuit 113, compressed in the unique mode again, and stored in the compression memory 111 again.

In particular, the amount of compressed image data for one screen is
When the capacity exceeds the capacity of the compression memory 111, it is recompressed so that one screen can be stored in the compression memory 111. The amount of the compressed image data is sent from the host computer 119 to the command interpreter 118.

(8) Image Reproduction When reproducing an image based on the compressed data stored in the compression memory 111, the decompressed image data is subjected to real-time processing corresponding to the recording speed of the color printer 102.

The restored image data sent to the double buffer 108 through expansion and scaling is converted by the selector 107 into a color space conversion circuit 106 to record color signals (Y, M, C,
K), and only a necessary portion is trimmed in the trimming circuit 104, so that the video interface 10
3 to the color printer 102.

On the other hand, a control command for controlling the start of recording and the like is sent through the video I / F 122.

In this embodiment, one set of photosensitive drums,
Since a multi-transfer type electrophotographic printer that repeats the transfer four times of Y, M, C, and K on the transfer drum is used,
It is necessary to sequentially send image data for one screen of C and K to the printer. Therefore, reading of four times for four colors from the compression memory 111 is repeated, and the color conversion circuit 106 reads Y, M, C,
Data for one screen is output in the order of K.

(9) Output to Computer The raw image data and the compressed image data can be output via the SCSI I / F 116.

The raw image data is sent from the double buffer 108 to the host computer 119 via the SCSI I / F 116.

On the other hand, in the case of compressed image data, the processing differs depending on whether or not the output destination computer accepts the compressed image data in the unique mode.

If the communication partner is a partner that accepts the unique mode, color image data with high image quality and high coding efficiency can be transmitted by transmitting the compressed data and the compression table data. However, this data cannot be transmitted if the other party supports only the standard mode JPEG system. Therefore, the data is transmitted while being converted into the standard JPEG format by the following method.

One page of data compressed in the unique mode and 8
Using 112 table information data determined from the information in the × 8 block and the compression memory progress information, the data is decompressed and restored by the decompression circuit 113 for each 8 × 8 block. The decompressed and restored data is stored in a double buffer memory 108 through a scaling circuit 114. The decompressed data is compressed in a compression table by a fixed table within one page and stored in 128 compression buffer memories.
Is transmitted as standard JPEG data to the host via SCSI. This compressed data complies with the JPEG standard because one page is compressed in a fixed table. However, since the table is fixed, the control for keeping the compression memory constant is ineffective, and writing and recording in a fixed one-page memory is not possible. However, in a computer having a hard disk such as a host computer, there is no problem even if the amount of one page memory slightly varies. As in the case of a semiconductor memory, when a memory amount is fixed and a memory overflow occurs, a slight problem remains.

In the above-described embodiment, the compression is performed for each of the surfaces r, g, and b in the standard color space.
^{(L *, a *, b} *), (Y, C r, C b), (Y, I,
For example, Q and the like, compression may be performed separately for luminance and chromaticity. In this case, it is desirable that the image area is determined based on the luminance signal.

The printer is not limited to the electrophotographic system, but may be an inkjet thermal transfer system. Further, a printer (bubble jet printer) using a head of a type that discharges droplets using film boiling due to thermal energy may be used.

The compression method is not limited to the ADCT method. As the compression parameter, a coefficient of the Huffman table may be changed in addition to the quantization parameter.

[0060]

As described above, according to the present invention, efficient coding for quantizing an image with a quantization table corresponding to the characteristics of each block can be preferentially performed, and regardless of the characteristics of each block. It becomes possible to cope with an encoding method such as the JPEG standard for quantization. In particular, the coded data obtained by the former coding method can be transmitted quickly, and the coded data obtained by the latter coding method can be transmitted by generating from the coded data obtained by the former coding method. It is. Furthermore, according to the present invention, it is noted that the former coding method is a method of changing a quantization table for each block, and in the former coding, quantization according to the characteristics of an image can be performed to some extent. At the same time, the amount of encoded data obtained here can be set to a desired amount.

[Brief description of the drawings]

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

FIG. 2 is a block diagram illustrating a configuration of a compression circuit.

FIG. 3 is a block diagram illustrating a configuration of an image area determination circuit.

[Explanation of symbols]

 110 compression circuit 111 compression memory 115 data amount controller 132 image area determination circuit

Continuation of the front page (56) References JP-A-62-208758 (JP, A) JP-A-3-166887 (JP, A) JP-A-2-70127 (JP, A) JP-A-2-264585 (JP) JP-A-2-262785 (JP, A) JP-A-60-64572 (JP, A) JP-A-63-40475 (JP, A) JP-A-2-2055982 (JP, A) 62-178071 (JP, A) JP-A-1-144778 (JP, A) JP-A-3-13065 (JP, A) JP-A-2-248180 (JP, A) JP-A-2-194734 (JP, A A) (58) Fields surveyed (Int. Cl. ⁷ , DB name) H04N ^7/ 24-7/68 H04N 1/41-1/419 JICST file (JOIS)

Claims (4)

(57) [Claims] An input unit for inputting image data; a dividing unit for dividing the image data into a plurality of blocks; a converting unit for performing frequency conversion on each block; and a determining unit for determining an image feature of each block. Encoding means for quantizing and encoding the frequency transform coefficient of each block obtained by the conversion means by selectively using a quantization table according to the determination result for each block to obtain first encoded data And data indicating the first coded data and the determination result depending on whether or not the transmission partner accepts the coded data quantized and coded based on the selective use of the quantization table for each block. Or transmit the second encoded data obtained by decoding the first encoded data and re-encoding each block using a single quantization table. Control means for controlling, wherein the determination means determines the determination result in accordance with an input amount of image data by the input means and the encoding means so as to set a code amount of the first encoded data to a predetermined code amount. An image processing apparatus for correcting the generated code amount according to a result of comparison. 2. The image processing apparatus according to claim 1, wherein the determination unit determines the degree of each of the blocks as a character or a photograph. 3. The image processing apparatus according to claim 1, wherein the frequency transform is DCT. 4. The image processing apparatus according to claim 1, wherein the encoding by said encoding means is Huffman encoding.