JP2702133B2 - Image processing method - Google Patents

Description

Description: BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method, and more particularly to an image processing method for performing image processing on a multilevel digital image. [Prior Art] In recent years, with the spread of facsimile communication or the development of high-definition printers such as laser printers, not only conventional line drawings such as characters and figures, but also characters such as newspapers and catalogs, images mixed with photographs, characters, There is an increasing demand for communication or output processing of halftone image mixed images. Generally, in this type of processing, binarization processing of an image is usually performed. Conventionally, when binarizing an image including a halftone, a binarization process called a dither method is performed to express a halftone. However, the dither method has the following disadvantages. (A) Moire is generated between a halftone dot and dither matrix for a halftone dot document, and the image is remarkably deteriorated. (B) Characters are cut by dither, and the quality is degraded. Therefore, there is a method in which a halftone portion is separated from a character and a thin line portion, and dither processing is performed in the former, and binarization processing of a fixed threshold is performed in the latter. For example, as shown in FIG. 8, P _{max is the} maximum density and P _min is the minimum density in an m × n block, and P _max −P _min > T. In the case of P _max −P _min ≦ T (3) A method of flattening → binarization by the dither method has been proposed. However, even if such a method is used, the aforementioned disadvantage (b)
Is improved, but the problem (a) still remains. In order to solve the problem (a), when it is determined that the image is flat, it is conceivable to perform the dithering after performing the smoothing process. It is easy to judge, and the halftone portion becomes a rough image due to binarization by a fixed threshold value. Further, if the condition that the halftone dot flat portion is determined as the flat portion is set, the edge of the character portion band is likely to be determined as the flat portion, and the problem (b) cannot be solved. [Problems to be Solved by the Invention] The present invention eliminates the above-mentioned drawbacks of the conventional example, and makes it possible to use various types of images, such as a document image, a halftone dot image, and a photographic image, without deteriorating the halftone. Is provided. In particular, a high-definition image is provided by emphasizing an edge portion of a character portion while removing moiré generated between a halftone dot and a dither matrix, which has been regarded as a defect in the related art, by smoothing. [Means for Solving the Problem] In order to solve this problem, an image processing method according to the present invention divides an image into blocks of a predetermined size, calculates an edge amount in a block of interest, and Calculate the edge amount between blocks with the surrounding blocks,
The feature of the image in the block of interest is determined based on the edge amount in the block of interest and the edge amount between the blocks. [Operation] In such a configuration, it is possible to express a halftone without deteriorating edges by determining features of an image in the block of interest based on the amount of edges in the block of interest and the amount of edges between blocks. it can. Embodiment An embodiment of the present invention will be described below in detail with reference to the accompanying drawings. FIG. 1 is a block diagram of the image processing apparatus according to the present embodiment. The image signal enters the line buffer 12 from the image input unit 11, and 4 × 4 pixels are regarded as one block, and the orthogonal transformation unit 13 performs orthogonal transformation of the block of interest. After that, the edge amount detector 14
At, the edge amount in the block is determined. On the other hand, the block average comparison unit 15 determines the association with the surrounding blocks. Here, a memory (not shown) for storing the average value of the previous block and the average value of the previous block is built in. The judgment result of the edge amount detection unit 14 is output in 2 bits, the judgment result of the block average comparison unit 15 is output in 1 bit, each signal is input to the selector 18, and one of the two output image data is selected. Is done. On the other hand, after the image data is stored in the line buffer 12, the image data is smoothed by the smoothing filter 16 and simultaneously edge-emphasized by the edge emphasizing unit 17, and two types of image data are input to the selector 18. Here, the edge amount detection unit
14, one of them is selected according to the output content of the block average comparison section 15. Output processing of the selected image data is performed by the output unit 19. First, a conversion process in the orthogonal transform unit 13 will be described.
In this embodiment, an example in which a block is divided into 4 × 4 and Hadamard transform is performed will be described. FIG. 2 shows the division into 4 × 4 blocks in this embodiment. 3 (a) and 3 (b) are conceptual diagrams of orthogonal transform (Hadamard transform in this embodiment), and FIG. 4 is a diagram showing a sequence component. The concept of the orthogonal transformation of the present embodiment will be described with reference to FIGS. 3 (a), (b) and FIG. In Figure 2, the target block X, the previous block C, and the same position block of the previous line B, and previous blocks of the previous line and _{A, P 1, ..., P} 16 in each image density of the target block X 8
Expressed in bits. Figure 3 (a) pixel data P ₁ of the _{4 × 4, P 2, ...} , and the P ₁₆ Hadamard transform, Y _11, Y ₁₂ of FIG. 3 (b), ..., the sequency components of Y ₄₄ Perform the conversion. In this embodiment, Y ₁₁ to Y ₄₄
Convert to a 10-bit sequence component. The Hadamard transform uses a Walsh type, and Y _{11 to} Y ₄₄ are sequence components at corresponding positions in FIG. The edge amount of the image data of one block is detected by the sequence components Y _{11 to} Y ₄₄ of one block (4 × 4 pixels) subjected to the Hadamard transform. Since information having a high correlation such as an image can be concentrated in a low frequency region by performing an orthogonal transformation, the characteristics of the image, for example, an edge portion and a non-edge portion are classified according to the magnitude of the power of the low-sequence component. be able to. Next, an example of an edge amount detection method of the edge amount detection unit 14 will be described in detail. In this embodiment, the edge amount is defined as shown in FIG. The edge amount as shown in FIG. 5 is defined as the sum of the absolute values of the components shown by hatching in the case of more than Sureshiyorudo there is edge amount E _D and edge. In the present embodiment, two levels of thresholds T ₁ and T ₂ (T ₁ <T ₂ ) are provided, and flat when E _D ≦ T ₁ and weak when T ₁ <E _D ≦ T ₂ ... (1) T ₂ < In case of E _{D It} is judged as strong edge. In this case, a flat portion of a photographic image and a background portion of a character image are determined to be a flat portion, a flat portion of a weak edge or a halftone dot is determined to be a weak edge portion, and a strong portion of a character portion and an edge of an image are determined to be a strong edge portion. T ₁ , T as
_I have decided ₂ . However, it is difficult to distinguish a weak edge such as a photograph edge from a halftone dot flat portion only by this determination. Therefore, the block average comparing section 15 compares the average values between the blocks as shown in FIG. In FIG. 6, Y _{11 to} Y ₄₄ indicate the values of the respective components of the block of interest X after the Hadamard transform, as in FIG. Y ₁₁ represents a value proportional to the average value of the concentration of block. Y _11a , Y _11b , and Y _11c are the average values of adjacent blocks as shown in FIG. In this embodiment, when the following equation is satisfied, it is determined that there is no edge between blocks. | Y ₁₁ −Y _11a | ≦ T ₄ and | Y ₁₁ −Y _11b | ≦ T ₄ … (2) and | Y ₁₁ −Y _11c | ≦ T ₄ where T ₄ is a predetermined threshold value. defining the _DEF ≦ T _4, if any of the above equation is not satisfied if there is edge between blocks, that is, the Y
_DEF> T is _four. High-definition halftone images (for example, 150 lines or 17
The flat portion (line 5) is determined to be an edge within a 4 × 4 block, but the average value between blocks does not change much. Therefore, the image can be classified as shown in FIG. 7 by using the edge judgment in the block and the edge judgment in the block. That is, after determining that the edge is flat, weak edge or strong edge by the determination of the expression (1), it is determined whether there is an edge between the blocks by the expression (2). According to the determination methods of the expressions (1) and (2), the classification shown in FIGS. Are as follows. Flat part: flat part in the block such as a photograph flat part, character image background ... flat part halftone dot adjacent to the edge of the character part or line drawing ... flat part of the halftone dot weak edge ... weak edge block in the photo etc. strong edge in the character: inside the character Strong edges such as complex patterns ... edges of characters etc. Among the above-mentioned classifications, it is preferable to regard as flat parts and as edge parts. In particular, when the halftone dot flat portion is regarded as an edge portion, it becomes a cause of generation of moire or becomes a rough image. Therefore, in this embodiment, by selecting the output of the smoothing process from the smoothing filter 16 for the flat portion and the edge-enhanced output from the edge emphasizing portion 17 for the edge portion, the character / thin line is selected. Does not cause deterioration, suppresses moiré due to halftone dots, and improves quality. In the present embodiment, classification is performed as shown in FIG. 7 based on the determinations made by equations (1) and (2), and smoothing processing or edge emphasis processing is performed according to the result. On the other hand, moiré occurs between the halftone dot and the dither matrix, which significantly deteriorates the image. (B) It is to solve the problems such as the character being divided by the dither and the quantile being lowered. In this embodiment, the Hadamard transform is performed in detecting the edge portion. However, the Hadamard transform may be performed by using the difference between the maximum density and the minimum density in the block as shown in FIG. 8 and equation (3). In this case, in FIG.
The edge amount detection unit 14 may perform the determination of Expression (3) without performing the Hadamard transform in 13. The block average comparison section 15 needs to calculate an average value. Thus, the embodiment can be implemented in the same manner as in FIG. Further, the edge amount is not limited to this example, and the edge amount detection unit may detect another edge amount such as Laplacian. As described above, according to the present embodiment, it is possible not only to improve the quality of the halftone image without deteriorating the characters and thin lines, but also to suppress the moire of the halftone dot image which has been a problem in the past. Natsuta [Effects of the Invention] According to the present invention, it is possible to provide an image processing method for expressing halftones of various types of images such as a document image, a halftone dot image, and a photographic image without deteriorating edges. In particular, it is possible to provide a high-quality image by emphasizing the edges of the character portion while smoothing the moire generated between the halftone dot and the dither matrix, which has been regarded as a defect, and smoothing the moire.

BRIEF DESCRIPTION OF THE DRAWINGS FIG. 1 is a block diagram of an image processing apparatus according to this embodiment, FIG. 2 is a diagram showing division into 4 × 4 blocks in this embodiment, and FIGS. b) is a conceptual diagram of the Hadamard transform of the present embodiment, FIG. 4 is a diagram showing a sequence component of the present embodiment, FIG. 5 is a diagram for explaining the definition of the edge amount of the present embodiment, and FIG. FIG. 7 is a diagram for explaining comparison of average values between blocks in the embodiment, FIG. 7 is a diagram for explaining classification of images in this embodiment, and FIG. 8 is a diagram based on a difference between a maximum density and a minimum density in a block. It is a figure explaining detection of an edge amount. In the figure, 11 ... an image input unit, 12 ... a line buffer, 13 ...
... Orthogonal transform unit, 14... Edge amount detection unit, 15... Block average comparison unit, 16... Smoothing filter, 17... Edge enhancement unit, 17, 18.

Claims (1)

(57) [Claims] The image is divided into blocks of a predetermined size, an edge amount in the block of interest is calculated, and an edge amount between blocks between the block of interest and peripheral blocks is calculated. An image processing method comprising: determining a feature of an image in the block of interest based on an edge amount between the blocks. 2. The edge amount in the target block is calculated based on image data in the target block, and the edge amount between the blocks is calculated based on image data of the target block and peripheral blocks. The image processing method according to claim 1, wherein