JPH0759086A - Picture processing unit - Google Patents

Abstract

PURPOSE:To suppress deterioration of picture quality of picture data and to attain efficient coding. CONSTITUTION:A character block discrimination hart 22 divides input color picture data into 8X8 picture element blocks and applies DCT transformation to obtain the absolute sum of 5 transformation coefficients belonging to a low frequency area among AC components of the transformation coefficients obtained through DCT transformation. When the sum is larger than a predetermined threshold level, the block is discriminated to be a character block and when it is smaller, the block is discriminated to be a natural picture block. Acoording to the result of discrimination, a linear quantization hart 24 selects an optimum coding table for an optimum quantization table by a Huffman coding hart 26 to execute quantization and coding.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing apparatus, and more particularly to an image processing apparatus for encoding image data.

[0002]

2. Description of the Related Art The JPEG system (Journal of the Institute of Image Electronics Engineers, Vol. 20, No. 1, No. 1, which is currently being internationally standardized is known as a highly efficient compression encoding system for color still images. 1991, pp50-58).

FIG. 12 is a block diagram showing an outline of a processing procedure of the JPEG system.

In the JPEG system, first, input image data is sampled and divided into blocks of 8 × 8 pixels,
Two-dimensional discrete cosine transform (hereinafter DC
(T) is performed (121 in FIG. 12). Next obtained 8
For the DCT transform coefficient of × 8, the quantization table (8 ×
Linear quantization is performed by using 8 pixels) with a different step size for each coefficient position (122 in FIG. 12). In the last quantized DCT transform coefficient, the DC coefficient is subjected to Huffman coding by taking the difference from the DC coefficient of the block of the same color component one before, and the AC coefficient is first subjected to zigzag scanning, One-dimensional conversion is performed, and entropy coding (2
(Dimensional Huffman encoding) is performed (123 in FIG. 12), and the obtained code is output as encoded data.

As an encoding method which is an extension of this method, the input image data is divided into a plurality of areas according to the characteristics of the image, such as a character portion and a natural image portion, and a quantization step size and a quantization are performed for each area. It is conceivable to use a different encoding method for each area, such as changing the table.

For example, input image data is subjected to DCT in block units of 8 × 8 pixels, the DCT transform coefficient obtained as a result is used to determine whether the block is a character block, and the DCT transform coefficient is quantized. The image processing apparatus is configured to perform quantization by switching the quantization table when performing the determination in accordance with the block determination result.

[0007]

However, in the above-mentioned conventional example, when an image in which images having different image characteristics such as characters and natural images are mixed is encoded, there is a problem that image deterioration occurs.

Further, in Huffman coding when such an image is divided into a plurality of areas and coding is performed for each area, the journal of the Television Society of Japan (Vol. 43, No. 12, 1989, pp1361-136) is used.
As shown in 9), since encoding was performed using one type of Huffman code table (optimized for the entire image regardless of JPEG recommendation and area) regardless of the area, Can not perform efficient Huffman coding,
There is a problem that the code amount becomes large.

The present invention has been made in view of the above-mentioned conventional example, and is capable of performing efficient encoding with little image quality deterioration even for images in which images having different image characteristics are mixed. An object is to provide a processing device.

[0010]

In order to achieve the above object, the image processing apparatus of the present invention has the following configuration. That is, an image processing device capable of inputting and encoding image data, wherein the input image data is divided into a plurality of blocks of a predetermined unit, and orthogonal transformation is performed for each of the plurality of blocks. Conversion means, discrimination means for discriminating image characteristics for each of a plurality of blocks using a transformation coefficient obtained from orthogonal transformation by the conversion means, and for each of the plurality of blocks according to a discrimination result by the discrimination means,
Quantizing means for quantizing the transform coefficient using an optimum quantization table, and quantization by the quantizing means using an optimum coding table for each of the plurality of blocks according to the discrimination result by the discriminating means. And an encoding unit that encodes the converted transform coefficient.

[0011]

According to the present invention having the above-described structure, the input image data is divided into a plurality of blocks, orthogonal transformation is performed for each block, and the image characteristic of each block is discriminated by the transformation coefficient obtained by the orthogonal transformation. According to the result, the quantization and coding are performed using the optimum quantization table and coding table for each of the plurality of blocks.

[0012]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENT A preferred embodiment of the present invention will be described in detail below with reference to the accompanying drawings.

FIG. 1 is a block diagram showing the outline of the overall configuration of an image processing apparatus that is a typical embodiment of the present invention.

In FIG. 1, 10 is an image input unit,
It is composed of an image reading device such as an image scanner including a CCD sensor, an interface of an external device such as a host computer, an SV camera, a video camera and the like. The image data input from the image input unit 10 is supplied to the input terminal 100 of the image storage unit 11 shown in FIG. Reference numeral 12 is an operation unit for the operator to specify an output destination of image data, 13 is an output control unit, the former is for selecting the output destination of the image data, and the latter is a synchronization signal for reading the memory (the printer engine unit together with the image output unit The ITOP signal or the like from the output control unit constituting the above, or output of connection information according to the image output unit (printer resolution) from the image output unit or by manual key input from the operation unit, for example. Reference numeral 125 denotes an output terminal of the image storage unit, and 126 denotes an input terminal of the synchronization signal of the image storage unit 11. 1
4 is an image display unit such as a display, 15 is a public line or L
A communication unit 16 that transmits and receives image data via the AN, for example, is an image output unit such as a laser beam printer that irradiates a laser beam on a photoconductor to form a latent image and visualizes the latent image.

The image output section 16 may be an ink jet printer, a thermal transfer printer, a dot printer or the like.

FIG. 2 is a block diagram showing a detailed structure of the image storage unit 11 shown in FIG. In the image storage unit 11,
The input image data is divided into blocks, and the image data of each block is encoded according to the JPEG method according to the image characteristics of each block.

In FIG. 2, reference numeral 21 denotes a DCT conversion unit that divides the input image data into blocks in units of 8 × 8 pixels, and performs DCT on each block, and 22 denotes a DCT obtained by the DCT conversion unit 21. A character block discriminating unit 23 for discriminating whether or not the block is a character block representing a character based on the conversion coefficient.
Quantization table switching units 23 and 24 that switch the quantization table according to the determination result of 2 are linear quantization units that linearly quantize DCT transform coefficients, 25 is a Huffman coding unit that Huffman codes quantized data, and 26 is a character. A region identification bit encoding unit that encodes the region identification bit that is the determination result of the block determination unit 22, and 27 is an image memory that stores encoded data.

Next, processing of the input image data in the image storage section 11 will be described.

First, the image data input from the input terminal 100 is divided into blocks in units of 8 × 8 pixels,
Two-dimensional discrete cosine transform (DCT) in the DCT transform unit 21
I do. The DCT transform coefficient obtained by the DCT transform unit 21 is composed of a direct current (DC) component and an alternating current (AC) component. The AC component is composed of various frequency components, which is represented as in FIG. In FIG. 3, a small square represents each DCT transform coefficient, and the lower left part represents the low frequency component and the lower right part represents the high frequency component. Further, the coefficient in the upper left shaded area represents a direct current (DC) component, and the other coefficients represent an alternating current (AC) component.

Now, in the character block discrimination unit 22, DC
Using the value of the conversion coefficient obtained by the T conversion unit 21, DC
It is determined whether the block subjected to T is a block representing a character (hereinafter referred to as a character block) or a block representing a natural image (hereinafter referred to as a natural image block).

In this embodiment, the character block is discriminated as follows.

(1) As shown in FIG. 4, the sum (S) of the absolute values of the five AC components in the low frequency region shown by the shaded areas among the DCT transform coefficients is obtained.

(2) This sum (S) is set to a predetermined threshold value (TH)
(Eg, 400). (3) If S ≧ TH, it is determined to be a character block, and if S <TH, it is determined to be a natural image block.

The result of the determination is as an area identification bit (if the value is "1", the block represents a character block, and if the value is "0", the block represents a natural image block). Is output. FIG. 5 is a visual representation of the relationship between area identification bits and blocks. As shown in FIG. 5, each block has one area identification bit indicating whether it is a character block or a natural image block.

The DCT transform coefficient is quantized by the linear quantizer 24.

As shown in FIG. 6 (b), the quantization table that has been generally used in the past has a quantization step size for the high frequency component of the DCT transform coefficient that is considerably larger than that for the low frequency component. Most of the high-frequency component values tend to be cut off by the processing. This is considered to be the cause of the image quality deterioration around the characters. Therefore, in the present embodiment, as the quantization table used for the block determined as the character block as shown in FIG. 6A, the high frequency component is finer and the low frequency component is slightly coarsely quantized. Use a quantization table. Then, for the natural image block, the quantization table shown in FIG. 6B is used.

The switching of the quantization table is performed by the quantization table switching unit 23 according to the determination result of the character block determination unit 22. By adaptively switching and using the two quantization tables in this way, it is possible to suppress the image quality deterioration of the character portion.

Using the quantization table thus obtained, the DCT transform coefficient is quantized by the linear quantization section 24, and the quantized data is further processed by the Huffman coding section 25.
Encode with. Encoded data (encoded data 1)
Are stored in the image memory 27.

7 to 8 are block diagrams showing the configuration of the Huffman coding unit 25. As shown in FIG. FIG. 7 shows the DCT transform coefficient D
FIG. 8 shows the configuration of a circuit for encoding the C component (DC coefficient), and FIG. 8 shows the configuration of a circuit for encoding the AC component (AC coefficient) of the DCT transform coefficient.

The values set in the DC Huffman code table 64 and the AC Huffman code table 75 shown in FIGS. 7 to 8 are set for each color component, DC component and AC component for each of the character block and the natural image block. The value to be set in each table may be obtained by performing optimization using the value in the block, or an appropriate Huffman code table may be given in advance for each block in consideration of the characteristics of the block and the like.

First, Huffman coding of DC coefficients will be described.

In this embodiment, the DC coefficient is first determined by the discrimination circuit 6
Input to 1. The discriminating circuit 61 discriminates whether the block is a character block or a natural image block according to the value of the area discriminating bit, and the adder 62 calculates the difference from the previous DC coefficient of the same color component and the same block type. The obtained difference value is divided into groups by the grouping circuit 63. The group number obtained by the grouping circuit 63 is encoded using the DC Huffman code table 64 selected according to the value of the area identification bit. The encoded data is output as a DC code. Further, the difference value within the group is directly output from the grouping circuit 63 as a DC additional bit.

Next, Huffman coding of AC coefficients will be described.

In the present embodiment, the AC coefficient is input to the zigzag scanning circuit 71, and zigzag scanning as shown in FIG. 9 is performed to convert it into a one-dimensional coefficient. The discrimination circuit 72 discriminates whether or not each one-dimensional coefficient is "0", and the length of a continuous "0" coefficient is counted by the run-length counter 73 and the run-length (NNN) is calculated.
N) is required. On the other hand, effective coefficients having a coefficient value other than "0" are grouped by the grouping circuit 74 in accordance with the value and assigned a group number. The group number (SSSS) and run length (NNNN) are input to the two-dimensional Huffman encoding circuit 76, and the group number (NNNN) and run length (SSS) as shown in FIG.
S) is encoded using the AC Huffman code table 75 selected according to the value of the area identification bit. The encoded data obtained as a result is output as an AC code. Further, the AC addition bit indicating the value of the AC coefficient in each group is directly output from the grouping circuit 74.

The Huffman code table defined for each block type is output as a header and is used at the time of decoding.

On the other hand, the area identification bits are also input to the area identification bit encoder 26, and entropy-encoded by using a binary encoding method (eg, MMR) using arithmetic codes. The resulting encoded data (encoded data 2) is output to the image memory 27. As a result, it is possible to suppress an increase in the code amount due to the generation of the area identification bit as much as possible.

Therefore, according to this embodiment, for each divided block of the input image data, it is determined whether the block is a character block or a natural image block, and according to the result of the determination, each block is divided into blocks. Since the quantization is performed using the optimum quantization table, it is possible to suppress the image quality deterioration due to the encoding. Further, in the Huffman coding process, since the Huffman coding is performed using the Huffman coding table in consideration of the block characteristics given for each block type, efficient coding can be performed.

In this embodiment, 8 × 8 pixels are used as the size of the block into which the input image data is divided, and DCT is used as the orthogonal transformation applied to the data of the block.
However, the present invention is not limited to this.
For example, a larger block or a smaller block may be used, or Hadamard transform or the like may be used as the orthogonal transform.

Further, in this embodiment, the sum of absolute values of the five AC components belonging to the low frequency region among the DCT transform coefficients is used for determining the character block, but the present invention is not limited to this. Absent. For example, the coefficient of another frequency component may be used, and the number of coefficients used may be other than five. Further, instead of the sum of absolute values, for example, the root mean square value may be used to obtain the index value. Further, the character block may be discriminated by applying, for example, a differential filter using the image data before the orthogonal transformation process.

Further, if the character area and the natural image area are structured and known in advance as shown in FIG. 11, they are divided into the a portion (character area) and the other area by manual input or some preprocessing. Separately, the apparatus can be configured to switch the quantization step for each region as in this embodiment.

Furthermore, in the present embodiment, the character of the image is distinguished between the character block and the natural image block, and the quantization table shown in FIG. Although the quantization is performed using the table, the present invention is not limited to this.
For example, another quantization table may be used, or one type of quantization table may be used to control the quantization width for each region. Also, as a way to distinguish images,
Instead of the natural image area, for example, a monochrome area,
You may distinguish by a color area etc.

The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of one device. Further, it goes without saying that the present invention can also be applied to the case of being achieved by supplying a program to a system or an apparatus.

[0043]

As described above, according to the present invention, the input image data is divided into a plurality of blocks, the orthogonal transformation is performed for each block, and the image characteristic of each block is determined by the transformation coefficient obtained by the orthogonal transformation. Discrimination is performed, and according to the discrimination result, quantization and encoding are performed using the optimum quantization table and encoding table for each of the plurality of blocks, so that encoding suitable for image characteristics is performed for each block. There is an effect that image quality deterioration due to encoding can be suppressed and efficient encoding can be performed.

[Brief description of drawings]

FIG. 1 is a block diagram showing an outline of the overall configuration of an image processing apparatus that is a representative embodiment of the present invention.

FIG. 2 is a block diagram showing a detailed configuration of an image storage unit 11 shown in FIG.

FIG. 3 is a diagram showing a configuration of a DCT transform coefficient.

FIG. 4 is a diagram showing DCT transform coefficients used for character block discrimination.

FIG. 5 is a diagram conceptually showing the relationship between area identification bits and blocks.

FIG. 6 is a diagram showing a structure of a quantization table.

FIG. 7 is a block diagram showing a configuration of a Huffman encoding unit that encodes a DC component of a DCT transform coefficient.

FIG. 8 is a block diagram showing a configuration of a Huffman encoding unit that encodes an AC component of a DCT transform coefficient.

FIG. 9 is a diagram showing an example of a zigzag scan of AC components.

FIG. 10 is a diagram showing a relationship between run lengths and group numbers.

FIG. 11 is a diagram showing an example of an image in which regions are structured in advance.

FIG. 12 is a diagram showing a flow of processing according to a conventional image compression encoding method.

[Explanation of symbols]

 21 DCT Transform Unit 22 Character Block Discrimination Unit 23 Quantization Table Switching Unit 24 Linear Quantization Unit 25 Huffman Encoding Unit 26 Region Identification Bit Encoding Unit 27 Image Memory

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI Technical indication location H03M 7/30 A 8522-5J 8837-5L G06F 15/70 330 F

Claims (3)

[Claims] 1. An image processing apparatus capable of inputting and encoding image data, comprising: dividing means for dividing the input image data into a plurality of blocks of a predetermined unit; and orthogonally for each of the plurality of blocks. Transforming means for transforming, discriminating means for discriminating image characteristics for each of the plurality of blocks using transform coefficients obtained from orthogonal transformation by the transforming means, and for each of the plurality of blocks in accordance with the discrimination result by the discriminating means A quantizing means for quantizing the transform coefficient using an optimum quantizing table; and, according to the discrimination result by the discriminating means, for each of the plurality of blocks, by the quantizing means using the optimal coding table. An image processing apparatus comprising: an encoding unit that encodes a quantized transform coefficient. 2. The image processing apparatus according to claim 1, wherein the image characteristic is determined by using a predetermined frequency component in the conversion coefficient. 3. The image processing apparatus according to claim 1, wherein the encoding means performs encoding by applying a Huffman encoding method.