JPH05143726A - Method for processing image and device therefor - Google Patents

Abstract

PURPOSE:To provide the image processing method and device therefor which can transform a binary image into a satisfactory multilevel image. CONSTITUTION:First of all, in a step S101, binary image data in the window of 7X7 containing an attention picture element are stored and in a step S102, an edge is detected and an edge direction is judged. Next, respective filters are selected corresponding to the judged result. Namely, in the case of half tone (a), a filter 1 is selected (step S103), in the case of a longitudinal edge (b), a filter 2 is selected (step S104), in the case of a lateral edge (c), a filter 3 is selected (step S105) and in the case of an inclined edge (d), a filter 4 is selected (step S106). In a step S107, the sum of products is calculated to the stored binary image data by the filters and the data are converted into multilevel data.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing method and apparatus for converting binary image data into high definition multivalued image data.

[0002]

2. Description of the Related Art Conventionally, a method of estimating multi-valued image data from binary image data sets a window including a target pixel to be converted, and takes a weighted average value of the binary image data existing in the window. It was done by the method, namely the smoothing filter.

However, in the image, there are high-frequency parts such as the contours and characters of the face and low-frequency parts such as the sky and the cheeks of the face. The high-frequency parts clearly convert the edges, and the low-frequency parts are 2 It is required to remove the graininess peculiar to the value image and perform smooth conversion. Therefore, a method of discriminating the edge portion and the halftone portion of the image and changing the multi-value conversion means for each region has been considered. That is, for the edge portion, a mountain-shaped filter in which the weight of the pixel of interest as shown in FIG. 13 is increased is used, and for the halftone portion, a smooth filter having a larger smoothing effect as in FIG. Used.

[0004]

However, in order to represent an image in binary, pseudo halftone densities are represented by dot intervals, and the rising and falling portions of the edges are as shown in FIG. Unevenness called a notch occurs. In particular, it has two characteristics of concentration preservation such as error diffusion method and concentration preservation method.
Notching is likely to occur in the quantification method. As an example, the notch of the character portion by binarization by the error diffusion method will be described. This error diffusion method will be briefly described as a binarization method in which an error generated by binarization by a threshold value is dispersed to unprocessed pixels located in the vicinity. More specifically, 2
When a 56-gradation multi-valued image is binarized by a certain threshold THRE, if the level of the pixel to be binarized is X, when X> THRE, the pixel is dot-printed. ..
That is, since it is converted into "255" in multi-valued level, the error that occurs is ERROR = X-255. Further, when X <THRE, it is converted to "0", so the error that occurs is ERROR = X.

Generally, as shown in FIG. 15 (a),
It is good to distribute the error ERROR in a large matrix, but here, for the sake of simplicity, description will be made using the matrix of (b). In the figure, * represents the pixel of interest, and the coefficient of the matrix represents the ratio of the errors distributed to that pixel. When the multi-valued image data shown in FIG. 16 is processed with the threshold value 127 by the error diffusion method of the matrix shown in FIG. 15 (b), the obtained binary image has notches as shown on the right side of FIG. Clearly the image.

Further, in the method of dividing an image into two regions and performing multi-value conversion for each region, since a mountain-shaped filter having a large center is used as a multi-value conversion method of a character part, multi-value conversion leaving a notch at an edge part is performed. Therefore, there is a drawback that the joint between the halftone portion and the edge portion looks unnatural.

The present invention has been made to solve the above problems, and an object of the present invention is to provide an image processing method and apparatus capable of converting a binary image into a good multi-valued image.

[0008]

In order to achieve the above object and to achieve the above object, the image processing apparatus of the present invention has the following configuration.

In an image processing apparatus for converting binary image data into multi-valued image data, a storage means for storing the binary image data in a window including a pixel of interest and a binary image data stored in the storage means are provided. It has edge detection means for detecting edges using a plurality of windows having different sizes, and multi-value conversion means for converting into multi-value image data according to the result of the edge detection means. Further preferably, the edge detecting means detects an edge and an edge direction by a first-order differential filter.

More preferably, the multi-value conversion means is provided with a plurality of filters according to the edge direction in the edge detection means, and a desired filter is selected and converted into multi-value image data. ..

Further preferably, the multi-value converting means is characterized in that the multi-value converting means converts the multi-value image data using a neural network.

Further, the image processing method of the present invention has the following steps.

In an image processing method for converting binary image data into multi-valued image data, binary image data in a window including a pixel of interest is stored, and an edge is detected based on the stored binary image data. The multi-valued image data is converted according to the edge detection result.

With the above structure, the binary image data in the window including the pixel of interest is stored, the edges are detected using a plurality of windows having different sizes based on the stored binary image data, and the detection result is obtained. It operates so as to be converted into multi-valued image data according to.

[0014]

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

FIG. 1 is a flow chart showing the processing procedure in this embodiment.

First, in step S101, binary image data in a 7 × 7 window including a target pixel is stored. This is because it is necessary to make a determination on a plane basis rather than a pixel basis in order to perform area determination of a binary image without erroneous determination. Further, in this embodiment, a large window of 7 × 7 is provided, but the present invention is not limited to this. Then, in step S102, edge detection and edge direction determination are performed. Specifically, 7 in (a) and (b) shown in FIG.
Double determination is performed for x7 and 5x5 primary differential filters of (c) and (d). In other words, in the present embodiment, the 7 × 7 filter can detect not only a sharp edge of a character or the like but also a relatively smooth edge existing in the halftone image, but conversely, the frequency is high in a high frequency portion that is large in the character portion. There is an edge that cannot be detected because it cancels out. So 5x
By using the filter of No. 5 and increasing the threshold value, a sharp high frequency portion can be detected as an edge.

For the purpose of only multi-value conversion with a sharp high frequency portion such as a character as an edge portion, a 5 × 5 first-order differential filter having a high threshold value is sufficient. In this embodiment, first, edge detection and edge direction determination are performed using a 5 × 5 filter. A flowchart of the processing is shown in FIG.

This process is performed by using the pixels contained in the 5 × 5 window centered on the pixel of interest among the 7 × 7 binary image data stored in step S101 described above. Here, the threshold THRE is set to "6". First, in step S301, the sum of products of the filter (c) shown in FIG. 2 and the binary image data is calculated to be an absolute value, and the vertical edge parameter sum1 is calculated. Next, in step S302, the lateral edge parameter sum2 is calculated in the same manner by the filter of (d) shown in FIG. Then, in step S303, it is determined whether the edge is a diagonal edge. As the judgment criterion, sum1 + sum2 is equal to or larger than the threshold value, and the difference between sum1 and sum2 is small. The threshold value of the difference is "4".

If the above conditions are met, the process proceeds to step S304, and the pixel is determined to be a diagonal edge. However, if it is not the case, the process proceeds to step S305, and it is checked whether or not the vertical edge, that is, sum1> THRE / 2. If YES here, step S
In step 306, the pixel is determined to be a vertical edge. However, if NO, the process proceeds to step S307 to check whether or not a horizontal edge, that is, sum2> THRE / 2, and if YES, the process proceeds to step S308 to determine the pixel as a horizontal edge. However, if NO, the process advances to step S309 to determine that the pixel is halftone.

For example, if the binary image data has a shape as shown in FIG.
um1 is “10”. On the other hand, the numerical value sum2 is "1"
Is. That is, since the difference between sum1 and sum2 is larger than the threshold value, the condition of step S303 is not met, but the condition of step S305 is met, and the pixel is determined to be a vertical edge. Further, in the case of the binary image data having the shape as shown in (b), sum1 is “1”,
sum2 is “8”. Therefore, it is determined to be a lateral edge. Furthermore, in binary image data like (c),
sum1 is “6” and sum2 is also “6”. That is,
Since sum1 + sum2 is "12", which exceeds the threshold THRE, and sum1-sum2 is "0", the condition of step S303 is satisfied. Therefore, it is determined to be a diagonal edge. As described above, the edge detection and the edge direction determination are performed in the 5 × 5 window size, and when it is determined that there is no edge, the edge determination is further performed in the 7 × 7 window size and the determination result is also taken into consideration. Multi-value conversion. By performing edge detection and edge direction determination using windows of different sizes in this way, it becomes easier to detect edges in the halftone portion, and the determination accuracy is significantly improved.

Next, the multi-value conversion filter in this embodiment will be described. FIG. 5 is a diagram showing four multi-value conversion filters. 5A is a filter for a halftone portion, FIG. 5B is a filter for diagonal edges, FIG. 5C is a filter for vertical edges, and FIG. 5D is a filter for horizontal edges. is there. First, the filter for the halftone portion of (a) shown in FIG. 5 is a gentle mountain-shaped filter.
It is better to have a little chevron because the harmonic components unique to the ED image can be erased. A filter having high-frequency components in both the vertical and horizontal directions is used for the diagonal edge filter of FIG.
Further, a filter having a high frequency component only in the vertical direction is used as the vertical edge filter of (c). Then, as the horizontal edge filter of the same (d), a filter having a high frequency component only in the horizontal direction is used.

Here, when the processing in step S102 is completed, the above-mentioned filter is selected according to the result. That is, if it is a halftone, filter 1 (step S103), if it is a vertical edge, filter 2 (step S104), if it is a horizontal edge, filter 3 (step S105), and if it is an oblique edge, filter 2 4 (step S106).
Next, in step S107, the binary image data stored in step S101 is summed with the above-mentioned filter and converted into multi-valued data. For example, when the edge detection and the edge direction determination are performed on the binary image data as shown in FIG. 6 by the method described above, the edge determination is performed as shown in FIG.
Further, when the multi-value conversion is performed by the above method, the binary image data of FIG. 6 is multi-value converted as shown in FIG. On the other hand, when the binary image data of FIG. 6 is processed only by the two judgments of the halftone and the edge portion by the conventional chevron filter, as shown in FIG. 10, the halftone portion and the edge portion where the notch remains are shown. The image becomes unnatural at the joint.

FIG. 10 is a schematic block diagram of the image processing apparatus in this embodiment. Here, the binary image data sent from the image input device (not shown) is input to the 7-line buffer 1101. This line buffer 110
In 1 is stored data of 7 lines which are continuous in the vertical direction of the image data. Then, the data from the left end to the right end at the same horizontal position among the 7 lines is sequentially input to the delay element (latch) group 1102 in synchronization with the clock.
In the latch group 1102, the data in the horizontal and vertical 7 × 7 small area is latched.

The latch group 1102 is composed of 49 latches, and the output of each latch is input to the edge detection and edge direction determination circuit 1103. The edge detection and edge direction determination circuit 1103 calculates from the input data,
Output judgment data. The judgment data is the selector 11
The control data is 08, and the selector 1108 selects one of the four filters 1104 to 1107 according to the control data. Then, based on the filter selected by the selector 1108, the multi-value conversion circuit 1109
The data stored in the latch group 1102 is converted into multi-valued data.

FIG. 11 is a detailed block diagram of the edge detection and edge determination circuit 1103 described above. here,
Since the data is binary, the first derivative filter can be calculated only by the adder circuit and the subtractor circuit. In the figure, 1201 is a part for calculating sum1 by the vertical direction filter,
Reference numeral 1202 is a part for calculating sum2 by the horizontal filter. Reference numeral 1203 denotes a portion for judging an oblique edge, which adds the data 1211 and 1212 and compares it with the threshold value THRE. On the other hand, data 1211 and 1212
Is judged by comparing the data ignoring the lower 2 bits of the two data. Then, the output of these outputs is ANDed to obtain the diagonal edge determination result.

Reference numeral 1204 denotes a portion for determining a vertical edge, which compares the data 1212 with THRE / 2. Similarly, 1205 is a portion for determining a horizontal edge,
The data 1212 is compared to THRE / 2.

In this way, the determination data 1213 indicating the edge determination result is output from the edge detection and edge determination circuit. FIG. 13 is a block diagram of the multi-value conversion circuit 1109. As shown in the figure, the coefficient of the selected filter and each pixel are multiplied by the multiplier 1301, and the next adder 1
The final multi-valued data 1 is added by 302
Converted to 303. In the case of a 7 × 7 window, the same configuration as in FIG. 11 can be used.

Since the edge detection and the edge determination are performed by the 5 × 5 and 7 × 7 first-order differential filters as described above, the edge determination can be performed accurately. The shape of the filter does not have to be square. Also,
The multi-value conversion may be performed by a neural network that matches the characteristics of the pixels. In particular, it is effective to use a neural network for the diagonal edge portion and the halftone portion.

Further, in the above-described embodiment, the edge judgment is divided into four, that is, the halftone, the slanted edge, the horizontal edge, and the vertical edge. As described above, a better image can be obtained by performing five edge determinations in total.

According to the embodiment described above, the halftone portion has no graininess, the edge portion has no notches, and the joint between the halftone and the edge portion can be smoothly converted. Multi-valued image data can be obtained. The present invention may be applied to a system including a plurality of devices or an apparatus including a single device. Further, it goes without saying that the present invention can be applied to the case where it is achieved by supplying a program to a system or an apparatus.

[0031]

As described above, according to the present invention,
It is possible to convert a binary image into a good multi-valued image.

[Brief description of drawings]

FIG. 1 is a flowchart showing a processing procedure in this embodiment.

FIG. 2 is a diagram showing a first-order differential filter used for edge detection and edge determination.

FIG. 3 is a flowchart showing edge detection and edge determination processing of FIG.

FIG. 4 is a diagram showing binary image data.

FIG. 5 is a diagram showing a filter used for multi-value conversion.

FIG. 6 is a diagram showing binary image data.

FIG. 7 is a diagram showing a result of edge detection and edge determination of the image of FIG.

FIG. 8 is a diagram showing a result of processing the binary image data of FIG. 6 in this embodiment.

9 is a diagram showing a result of processing the binary image data of FIG. 6 by a conventional method.

FIG. 10 is a schematic block diagram of an image processing apparatus in this embodiment.

11 is a block diagram of the edge detection and edge determination circuit of FIG.

12 is a block diagram of the multi-value conversion circuit of FIG.

FIG. 13 is a diagram showing two types of conventional multi-value conversion filters.

FIG. 14 is a diagram illustrating a notch.

FIG. 15 is a diagram showing error diffusion parameters of the error diffusion method.

FIG. 16 is a diagram showing how a multi-valued image is binarized by an error diffusion method.

Claims (5)

[Claims] 1. An image processing apparatus for converting binary image data into multi-valued image data, a storage means for storing the binary image data in a window including a pixel of interest, and a binary image stored in the storage means. An edge detecting means for detecting an edge by using a plurality of windows having different sizes based on the data; and a multi-value converting means for converting into multi-value image data according to a result of the edge detecting means. Image processing device. 2. The image processing apparatus according to claim 1, wherein the edge detecting unit detects an edge and an edge direction by a first-order differential filter. 3. The multi-value converting means comprises a plurality of filters according to the edge direction in the edge detecting means, and selects a desired filter to convert the multi-valued image data. 1. The image processing device according to 1. 4. The image processing apparatus according to claim 1, wherein the multi-value conversion unit converts the multi-value image data using a neural network. 5. An image processing method for converting binary image data into multi-valued image data, wherein binary image data in a window including a pixel of interest is stored and edges are stored based on the stored binary image data. An image processing method, which comprises detecting and converting into multivalued image data according to the edge detection result.