JPH07121699A - Image processing device - Google Patents

Abstract

PURPOSE:To make only required information into a vector by automatically sampling a texture area and the contour(edge) of it from image data, deleting the texture area from the image data, and thinning the contour of the texture area. CONSTITUTION:This device is constituted of an image input part 101, an arithmetic part 102, memory 103, an image output part 104, an image display part 105, and a pointing device 106. The texture area and the contour(edge) of it are automatically sampled from the image data, and the texture area is deleted from the image data, then, the contour of the texture area is thinned. Also, the texture area is made into the vector as a polygon consisting of uniform texture images as it is, and it is made into image descriptive language so as to possess the kind of the image in a closed area as attribute. Furthermore, a paint-out area and the contourpart of it in the image data are automatically sampled, and a sampled paint-out area is deleted from the image data, then, a paintout contour part is thinned.

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 for processing raster data obtained by reading a general document image with a scanner or the like or raster data of an image drawn on a computer. More specifically, the present invention relates to an image processing apparatus that recognizes a texture area or a filled area that is often used in an office document or a drawing as a closed area and deletes the closed area to enable efficient vectorization. .

[0002]

2. Description of the Related Art A general image is composed of elements such as text, line graphics, uniform areas, and natural images such as photographs. As described above, when a general image including a plurality of types of constituent elements is processed on a computer, it is necessary to perform processing according to each area. Conventionally, there are the following examples to meet such a demand. In this example, when there is a filled area in the line image, different processing is performed on the filled area and the area of the line image. The details will be described below. FIG. 29 shows Japanese Patent Laid-Open No. 1-20
FIG. 6 is a block diagram of a conventional graphic processing device disclosed in Japanese Patent No. 6470. The scanner 1 reads a figure drawn on paper and outputs a binary image signal, and is, for example, an image scanner or a television camera. The binary image signal output from the scanner 1 is given to the CPU 2. CPU2
A keyboard 3, an image memory 4 and a vector data memory 5 are connected to. The keyboard 3 is used to specify a line figure and a filled figure in the image read by the scanner 1 and to input the number of times for thinning. The CPU 2 stores the binary image data read by the scanner 1 in the image memory 4 based on a built-in program, and based on the binary image data, thinning, edge (outline) extraction, vector Processing such as conversion. Vector data memory 5 is CP
Store the vectorized data by U2.

Next, the operation will be described with reference to FIGS. 30 and 31. FIG. 30 is a flow chart showing the operation of the graphic processing device of FIG. First, an image is read by the scanner 1, and its binary image data is given to the CPU 2. This binary image data is displayed on a display device (not shown) such as a monitor. At this time, the operator (person who operates the graphic processing device) determines whether the read image is a filled graphic, a line graphic, or an image in which the filled graphic and the line graphic are mixed. Figure 31
In the case of the filled figure as shown in (a), the operator designates the filled figure with the keyboard 3 and inputs the number of primary thinning. The CPU 2 carries out the thinning process a specified number of times according to the input number. Specifically, the pixels at the widthwise end of the filled figure shown in FIG.
Each pixel is deleted, and the result is shown in FIG. Next, the CPU 2 performs an edge extraction process for the primary thinned filled figure as shown in FIG. As a result, FIG.
Process to a graphic represented by only a relatively thick line as shown in (c). Next, the CPU 2 performs thinning processing again on the figure whose edge has been extracted. The thick line graphic in FIG. 31 (c) is processed into a thin line graphic as shown in FIG. This is called secondary thinning. Next, the CPU 2 converts the thinned line figure into vector data and stores it in the vector data memory 5. The above processing is performed on the filled area. If it is a line figure, only the secondary thinning is performed, and the line figure is processed so as to form a connection for one dot of a pixel. If the filled area and the line figure are mixed, the above-described processing is performed on the filled figure by designating the filled figure and the line figure with the keyboard 3, and only the edge is displayed as shown in FIG. 31 (d). It becomes the figure shown. Also, a line figure can be processed into a figure represented by a line figure consisting of a series of one pixel.

[0004]

Since the conventional graphic processing apparatus is configured as described above, a graphic including a filled area can be filled by extracting the edge of the filled area first. Vectorization is also possible for regions. However, the operator needs to specify the filled area on the displayed input image and input the number of times the primary thinning is performed, which is troublesome.
In addition to the filled area, if there is a hatching or dot area as in the image shown in FIG. 7A, all regular patterns that are meaningless to vectorize are thinned and vectorized, For example, it is not recognized as a hatched rectangle. Further, when processing various types of images, texture pattern data and color image color data are very effective data for expressing images. However, there is no conventional image processing apparatus that automatically converts such image information into an effective vector data form.

The present invention has been made to solve the above problems, and an object thereof is to process not only line images such as drawings but also various types of images.
Furthermore, the purpose is to convert a general image into an image description language convenient for editing on a computer.

[0006]

An image processing apparatus according to the present invention is an image processing apparatus that handles raster data, and has a texture region such as hatching (a region where brightness and color change are uniform is called a texture region). And means for automatically extracting the contour (edge) thereof, means for deleting the extracted texture region from the image data, and means for thinning the contour of the texture region.

Further, the image processing apparatus according to the present invention is an image processing apparatus which handles raster data, and a texture area such as hatching (an area having a uniform change in brightness and color is called a texture area) and its contour ( Edge) and the means for deleting the extracted texture area from the image data, and the means for vectorizing the outline of the thinned texture area as a polygon. An image description language conversion means is provided so that the type of the image in the area has an attribute.

Further, the image processing apparatus according to the present invention is, in an image processing apparatus which handles raster data, means for automatically extracting a filled area and its contour portion in image data, and an image of the extracted filled area. And means for thinning the outline of the filled area from the data.

Further, the image processing apparatus according to the present invention is, in an image processing apparatus which handles raster data, means for automatically extracting a filled area and its contour portion in the image data and an image of the extracted filled area. In addition to the method of deleting from the data and thinning the outline of the filled area, the method of vectorizing the outline of the thinned filled area as a polygon, and the attribute of the type of image in the closed area And a means for converting the image into a description language.

The image processing apparatus according to the present invention further comprises means for encoding the colors forming the image, means for reconstructing the image data using the color code, and vectorizing the encoded image. And a means for converting the color code into an image description language as an attribute of vector data.

Further, the image processing apparatus according to the present invention comprises means for coding the colors forming the image, means for reconstructing the image data using the code of this color, and a filled area from the coded image. A means for automatically extracting and a means for converting the extracted color code of the filled area into an image description language as an attribute of the filled area are provided.

Further, the image processing apparatus according to the present invention comprises means for encoding colors forming an image, means for reconstructing image data using the color codes, and texture areas from the encoded images. A means for automatically extracting and a means for converting the extracted color code of the texture area into an image description language as an attribute of the texture area are provided.

[0013]

In the present invention, the texture region and its contour (edge) are automatically extracted from the image data, the extracted texture region is deleted from the image data, and the contour of the texture region is thinned.

In the present invention, the texture region is vectorized as it is as a polygon having a uniform texture image, and the image description language is created so as to have the type of the image in the closed region as an attribute.

In the present invention, the filled area and its contour portion in the image data are automatically extracted, the extracted filled area is deleted from the image data, and the contour portion of the filled area is thinned.

In the present invention, the filled area is vectorized as it is as a polygon whose inside is filled,
The image description language is created so that the type of the image in the closed area has an attribute.

In the present invention, the colors forming the image are coded, the image data is reconstructed using this color code, the image represented by the color code is vectorized, and this color code is converted into vector data. Image description language is used as an attribute.

In the present invention, the colors forming the image are coded, the image data is reconstructed using the code of this color, the texture region is automatically extracted from the image expressed by the color code, and this extraction is performed. The code of the color of the generated texture area is converted into an image description language as an attribute of the texture area.

In the present invention, the color forming the image is coded, the image data is reconstructed using the code of this color, the filled area is automatically extracted from the image expressed by the color code, and this extraction is performed. The code of the color of the filled area is converted into an image description language as an attribute of the filled area.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described in detail below with reference to the drawings. FIG. 1 is a block diagram showing the outline of this image processing apparatus. The image processing apparatus includes an image input unit 101 and a calculation unit 10.
2. The memory 103, the image output unit 104, the image display unit 105, and the pointing device 106. Reference numeral 101 is an image input unit such as a color scanner. The calculation unit 102 processes the raster data of the bitmap image input from the image input unit. Reference numeral 103 is a memory for storing data such as intermediate results required for each process. Image output unit 104
Outputs the processed image data. An image display unit 105 displays the input target image of the process. The operator uses the pointing device 106 such as a mouse to specify a character area in the image on the image display unit 105. The data of the designated character area is cut off only for that portion, input to the OCR 107, and converted into a character code there.

When image data is input from the image input unit 101, an image processing mode setting screen as shown in FIG. 2 is displayed on the image display unit 105. The input image is displayed in the input image display area 201 on the screen. The image processing mode setting screen of FIG. 2 has an input image display area 201, a character area setting button 202, a contour mode button 203, an all image mode button 204, a color mode button 205, and a mono mode button 20.
6. A processing start button 207 is arranged. On the image processing mode setting screen, the operator selects the contour mode button 203 and the all image mode button 204 in the mode selection column.
The image processing mode is selected by clicking either one of them, and either the color mode button 205 or the mono mode button 206. Here, four modes are provided as the image processing mode. Regarding the two modes of the color mode and the mono mode, there are provided a contour mode in which only contour data is converted into an image description language, and an entire image mode in which an image in a closed region is converted into an image description language.

If there is a character area in the displayed input image, the operator previously sets the character area setting button 20.
2 Click (use the mouse to move the pointer over the button and press the button attached to the mouse),
Transition to the character area setting mode. FIG. 3 is an image processing mode setting screen when the mode is changed to the character area setting mode.
On the display screen of the character area setting mode of FIG. 3, the split decision button 301, the setting cancel button 302, the mode end button 30
3 are arranged. In the state of transition to the character area setting mode, the operator clicks the upper left and lower right points of the character area on the image displayed on the image display unit 201 to select the character area as a rectangle. At this time, as shown in the title of the image in FIG. 3, the selected area is "
As you can see that it has been selected, the upper left and lower right vertices are marked with dashed rectangles with white circles. After selecting the character area in this way, click the division decision button 102 stores the coordinates of the upper left and lower right points of the rectangle in the memory as the character area surrounded by the broken line rectangle with the white circles. As shown in the description at the bottom of the image, "Recognized as a character area" is marked with a dashed rectangle with black circles at the top left and bottom right vertices. When 303 is clicked, the arithmetic unit 102 cuts raster data in the character area and inputs it to the OCR 107.
And the coordinates of the upper left and lower right points of the rectangle of the character area are stored in the memory 103 as character area data.
The original image data with the character area deleted is stored in the memory 103 as graphic data. FIG. 4 shows respective image data before and after the above character area division processing. 4A is an original image, and FIGS. 4B and 4C are character area data (b) and graphic area data (c) obtained by the character area dividing process, respectively.

As described above, the character area is divided, the image processing mode is set, and when the operator clicks the processing start button 207, the image processing is executed according to the flow of FIG. 5, and the raster data of the image is finalized. Is converted into an image description language. First, for convenience, the definitions of variables and symbols used in this description will be described with reference to FIG. The coordinates of the section are (i, j) and the coordinates of the pixel are (x, y). The directions of the coordinate axes are as shown in FIG. In addition, the size of the section
Let m be the coordinate in the section be (x, y). Further, the lowest right section coordinate is (imax, jmax) and the pixel coordinate is (xmax, ymax). In addition, the variables of the image data handled in this processing are listed below. Image raster data: A (x, y) Fill area raster data: B (x, y) Edge area raster data: D (x, y) Texture area raster data: C (x, y) Raster data of the edge portion of the texture area: E (x, y) A (x, y) is set to 0 when the pixel at the coordinates (x, y) is white and is set to 1 when it is black. In the color mode, an integer of 2 or more is assigned as a color code to colors other than white and black. In addition, data B (x, y), C (x, y), D (x, y), and
The values of E (x, y) are all area numbers, and are 0 for pixels that are not extracted as areas. The outline of the processing will be described below with reference to the flowchart of FIG. First, the selection of the mono mode and the color mode is determined. Hereinafter, first, the mono mode processing will be described in detail with reference to FIG. 7. FIG. 7 is a diagram showing a process of processing image data in the image processing apparatus according to the embodiment of the present invention. Figure 7
5 includes image data A (x, y) obtained by the processing of FIG. 5 and B (x, y), C (x, y), D (x, y), E (x, y) formed during the processing. It is shown. 7 (a) to (g), each data is 0
Pixels are displayed in white, and one or more pixels are displayed in black.

FIG. 7A shows the image data A (x, y) of the graphic area in which the character area is deleted. In the preprocessing, the image is
The image is divided into sections as shown in (b), and the number of black pixels in each section is counted from the image data A (x, y). This parcel coordinate (i,
The number of black pixels in j) is expressed as pixels (i, j, code_black). Since the number of black pixels for each section is necessary for the subsequent processing, it is stored in the memory 103. Next, according to an algorithm described later, the filled area and its edge portion are extracted, and the filled areas are numbered in the order of extraction. This is called a region number and is represented by a variable K. The area number is the raster data B (x, y) of the filled area and the raster data D (x, y) of the edge of the filled area.
set to k B of the pixels that are not extracted as the filled area
The values of (x, y) and D (x, y) are 0. B (x, y) = k (1) D (x, y) = k (2) Here, the edge part in the filled area is the outermost pixel in the filled area. Next, the filled area and the edge portion are deleted from A (x, y). Specifically B (x, y)
The value of A (x, y) of a pixel that is not = 0 or D (x, y) = 0 is set to 0.
FIG. 7 (c) shows A (x, y) with the filled area deleted. Further, the filled area B (x, y) extracted here and its edge portion D (x, y) are shown together in FIG. 7 (d). Figure 7
In (d), the numeral is the area number k.

Next, the texture region and its edge portion are extracted according to an algorithm described later. The edge of the texture area is a closed loop that surrounds the texture area.
Similarly to the filled area, the area number is assigned to the texture area in the order of extraction. Then, the values of the raster data C (x, y) of the texture area and the raster data E (x, y) of the edge portion of the texture area are set to the area number k. Values of C (x, y) and E (x, y) of pixels that are not extracted as texture areas are 0
Is. C (x, y) = k (3) E (x, y) = k (4) At this time, as shown in FIG. 8, a code indicating the area number of the texture area and the type of texture image (hereinafter, this The code is referred to as a texture code) and stored in the memory 103. The texture code is registered with the texture image when the texture region is extracted and the texture image is of a new type. As shown in FIG. 9, raster data of a texture image of a predetermined size and the image feature amount of the texture image are stored in the memory 103 in association with the code. The image feature amount will be described later in detail. Next, the texture area and its edge portion are deleted. Similar to the filled area, C (x, y) = 0 or E (x, y) = 0
Set the value of A (x, y) in the non-texture area to 0. A (x, y) from which the texture area is deleted is shown in FIG. 7 (e).
The texture area C (x, y) and the edge portion E (x, y) extracted here are shown in FIG. 7 (f). In FIG. 7 (f), the numeral is the area number of the texture area.

After deleting the filled area and the texture area, line image processing is next performed. This target is A
The data is (x, y), D (x, y), and E (x, y). These data are summarized and shown in FIG. 7 (g). Here, since there are known techniques for thinning and vector conversion of line image processing to be handled,
I won't go into the details here. First, A (x, y), D (x,
y) and E (x, y) are thinned, and then A (x, y) is vectorized. Next, D (x, y) and E (x, y) are vectorized with the recognition that they are closed regions. Where D
Although (x, y) and E (x, y) are area number data, in a series of line image processing, if the value is 1 or more, the value is 1, and if 0, the value is 0. At that time, different processing is performed in the contour mode and the whole image mode. In the contour mode, vector conversion is simply performed. On the other hand, in the case of the all image mode, after the vector conversion of the edge image data of D (x, y) and E (x, y), refer to the texture code shown in FIG. Then, a texture code or a code indicating that the image is filled is added to the final image description language as an attribute of vector data. In this way, when vectorization is completed for A (x, y), D (x, y), and E (x, y), as shown by the black circles in the figure of FIG. The passing point is required.

After the above processing is completed, when the final image description language is formed, for example, the data shown in FIGS. 10 (a) and 10 (b) are obtained. FIG. 10 shows an image description language finally output by the image processing according to the embodiment of the present invention as a mono-contour mode.
It is the figure which divided and showed in the case of (a) and the mono all image mode (b). In all the image description languages in FIG. 10, Polg represents a polygon, Line represents a polygonal line, and the x and y coordinate values are arranged after those symbols. In the case of a polygonal line, these coordinates are the coordinates of two end points or passing points, and are described in the order of passing. When the x and y coordinates following Line are only data of two points, they are not line segments but simply line segments. On the other hand, in the case of a polygon, the coordinates of the vertices are described. In the contour mode image description language of FIG. 10A, the polygonal code Po
After lg, there is no code representing the image in the closed area, but in the image description language in the all image mode of FIG. 10B, the code representing the image in the closed area is shown. They are beta1 for the filled area and p for the texture area, respectively.
Codes of attern1 and pattern2 are added. The last number in this code corresponds to the texture code.

Each processing described above will be described in detail below. First, a method of extracting a filled area will be described. FIG. 11 is a flowchart of the processing for extracting the filled area. In the figure, k is a region number. First, the filled area is extracted with the section divided in the preprocessing as one unit. FIG. 12 shows a flow of processing for extracting a filled area in units of sections. Next, the filled area is extracted in pixel units by processing the area around the filled area. After the extraction of the filled area is completed in this way, the edge portion is extracted in the filled area.

When there is a filled area as shown in FIG. 12 (a), first, the main scan is the scan from the upper left section to the right direction, and for each section, whether or not it is a filled section is determined. The raster scan is performed while making the judgment. This judgment is whether or not the condition represented by the following expression (5) is satisfied. In equation (5), a is an optimized threshold value. pixels (i, j, code_black)> a After (5), the partition that satisfies Eq. (5) is called the filled partition. As shown in FIG. 12B, when one section of the filled area is found, the section-based continuation process is started from the section (i, j). In addition, labeling is performed with the region number k that is incremented by 1 by using the segment as a continuation start segment. This time,
The target of labeling is a section that is not yet labeled as a filled area. Therefore, one section should not belong to a plurality of filled areas. Since the initial value of k is 0, the area number of the first area is 1.

FIG. 14 shows a flowchart of the continuation process. By this continuation processing, continuous filled sections are labeled with the same value and recognized as one continuous area. The continuation process is performed by repeating the peripheral scanning of the filled section. The surrounding scan is performed by scanning the eight surrounding areas in the clockwise direction in the order shown in FIG. 13 and determining whether the area is a filled area similar to the continuation start area. The coordinates of the eight surrounding sections are represented by p and q in the flowchart of FIG. When the scanned surrounding section is a filled section, labeling is performed with the same value of the area number k as the continuous start section. In addition, the coordinate of the section labeled with the same label is a variable i that has num, which is the order of labeling, in an array.
It is stored in (num) and j (num). At this time, the continuation start section is not counted in the order in which the labels are given. 12
In (c), the peripheral scanning for the continuous start section is finished, and it is determined that the three sections, the right, the lower right, and the bottom are the filled sections, and the serial numbers num are numbered 1, 2, and 3. . Two variables i (num) with their serial numbers in the array,
The i coordinate and the j coordinate are stored in j (num). In FIG. 12C, i (1) = 2 i (2) = 2 i (3) = 1 j (1) = 1 j (2) = 2 j (3) = 2. When the surrounding scan is completed for the continuation start section,
The parameter indicating the number of sections for which the peripheral scanning is completed is set to 1. Now, this parameter is represented by the variable p_fin, and p_fin = 1. The parameter p_fin is also meant to represent the number of the section to be scanned next. Then change the value of p_fin to i (p
_fin) and j (p_fin), and the surrounding scan is performed on the section of the i coordinate and j coordinate obtained there. That is, FIG. 12 (c)
In, the peripheral scanning is performed for one section. The result is shown in FIG. 12 (d), and numbers 4 and 5 are newly added. At this time, a new labeling is not performed on a section that has already been labeled as a filled section. The values of the i coordinate and the j coordinate of those sections are stored as i (4) = 3 i (5) = 3 j (4) = 1 j (5) = 2, and the value of p_fin is incremented by 1.

This process is repeated until the peripheral scanning is finally completed for all the filled sections labeled with k. In the case of FIG. 12 (d), from 2 to 5 sections,
Although there is no new section to be labeled, the peripheral scanning is performed up to 5, and the section continuation process ends there (FIG. 12 (e)). Completion of partitioning is determined by comparing the value of the number of peripherally scanned partitions p_fin with the number of labeled partitions num.

When the continuation process from one continuation start section is completed, the raster scan is restarted from the section next to the continuation start section to search for the next continuation start section.
As shown in FIG. 12 (f), when a filled section is found again, the section is set as a continuation start section, and the continuation process is started from that section. At this time, the value of the area number k assigned to the filled section is obtained by incrementing the value of the area number k extracted immediately before by one. As described above, the label value is changed for each area, the filled area is extracted in units of partitions, and when the raster scan is completed up to the bottom-right partition, the pixels in the partition extracted as the filled area up to that point For, regarding the data B (x, y) of the filled area, B (x, y) = k. In FIG. 12 (g), in the actual filled area,
The portions surrounded by white lines are the two areas extracted in units of sections.

After the extraction of the filled area in units of sections is completed, the remaining portion of the filled area that cannot be extracted in units of sections in each area in the processing around the filled section in the flowchart of FIG. And edge part extraction processing is performed. This process scans the continuity of black pixels outward, as indicated by the arrows around the k = 1 filled section in FIG. 15 (a). The details will be described below with reference to the flowcharts of FIGS. FIG. 16 is a flowchart of the surrounding processing of the filled section in the image processing of the embodiment of the present invention. FIG. 17 is a flowchart of the intra-section continuous scanning process in FIG. In the flowchart, k is a variable indicating the area number, and k
max represents the number of areas. The surrounding processing is performed for each area. A raster scan is performed for each partition, and the continuity of black pixels from an adjacent painted partition is scanned for a partition that is not a painted area. In FIG. 15 (b), the center section is the not-filled section of interest. In the intra-section continuous scanning, first, it is determined whether or not the surrounding four sections on the upper, lower, left, and right sides are the painted area of the number k of interest. When there is a filled area in the four surrounding sections, the continuity of black pixels is determined from the surrounding filled section in the direction of the arrow shown in FIG. 15 (b). This corresponds to vertical processing in the flowchart of FIG. Further, as shown in FIG. 15 (c), when the diagonally adjacent sections at the upper right, lower right, upper left, and lower left are filled, and the upper, lower, left, and right sections sandwiching the diagonally adjacent section are not filled areas, Scan the continuity of the black pixels from the partition. This corresponds to the diagonal processing in the flowchart of FIG.

With reference to FIGS. 15 (d) to 15 (h), the pixel-based area extraction and edge portion extraction processing by the vertical processing and diagonal processing will be described. The process described here is a case where the continuity of black pixels is determined from the upper adjacent filled section as in the left end of FIG. Here, the vertical processing in the mono mode will be described. In the vertical processing, continuity from the outermost black pixel column of the filled section is scanned in the direction perpendicular to the black pixel column. In the diagonal processing, the continuity from the vertex adjacent to the target section in the filled section is evaluated. Figure 1
In 5 (d) to (h), one section shows one pixel, and the black section and the hatched section show black pixels. The hatching shows the two lines that are currently focused. The black pixel in the uppermost row indicates the black pixel row in the lowermost row in the upper adjacent filled section. In the case of vertical processing, the continuity of black pixels from one column facing the target partition is determined. In the case of diagonal processing, the continuity of black pixels from the vertex adjacent to the target section is determined.

The continuity of black pixels is judged by whether or not two adjacent lines overlap each other, and by repeating this, the filled area is extracted. This will be described step by step. FIG. 15 (d) shows the outermost one line of the filled area and the adjacent one line further outside, which are the first noticed lines of this processing. First, in the inner one line, the x coordinate x_in1 of the start point and the x coordinate x_in2 of the end point of the black pixel row are obtained. Next, the x coordinate x_out1 of the start point and the x coordinate x_ou of the end point of the outer black pixel row
Find t2. If x_out2> x_in1 or x_in2> x_out1) (6) is judged in these two lines and either of the above two expressions is satisfied, the continuity is confirmed by overlapping these two lines. And
Since it is confirmed that the outer black pixel row is a filled area, B (x, y) = k (k; area number).
Next, as shown in FIG. 15 (e), in order to confirm the continuity of the line further outside by one pixel, the x-coordinates of the start point and the end point of the outer black pixel row are newly set as the coordinates of the inner black pixel row. To do. x_in1 = x_out1, x_in2 = x_out2 (7) Similarly, until a new black pixel row is not extracted to the outside, or the extracted black pixel row does not overlap with the inner black pixel row for one pixel at all, the above The process of is repeated. With the above processing, the filled area can be extracted.

When the extraction of the filled area is completed, edge extraction processing is performed. This time, as shown by the arrows in FIGS. 15 (g) and 15 (h), the edge portion is extracted by searching for a place where the value of B (x, y) is changed in two directions of x and y. 15 (h), (i)
As shown in, the pixel represented by the dot pattern is an edge pixel, and this is D (x, y) = k. This is the end of the extraction processing of the filled area and its edge portion.

Next, a method of extracting a texture area will be described. FIG. 18 is a flowchart showing the texture region extraction processing. FIG. 18 is a flowchart of texture region extraction in the image processing of the embodiment of the present invention. In the figure, k is a region number. The process of extracting the texture region is basically the same as that for the filled region. First, the texture region is extracted in the unit of the divided sections in the pre-processing, and then the region of the pixel unit is extracted in the processing around the texture section.

The process of extracting a texture area in units of sections will be described. First, a raster scan is performed for each section while determining whether the section of interest is a texture image (determination 1). When one section of the texture image is extracted,
From that point, the continuation of the texture section is started. This time,
Judgment whether the image of the extracted texture section is equal to the texture image already coded (decision 2)
do. If they are equal, as shown in FIG. 8, the code and the area number are stored in association with each other. Otherwise, FIG.
As shown in, the predetermined size of the newly extracted texture image and the image feature amount of the texture image are stored in association with a new code, and then the new code is stored in association with this area number. To do. The texture segment continuation process is shown in the flowchart of FIG. This is
This is almost the same as the continuous processing of the filled area extraction shown in 3 and 14. In this processing, the difference from the extraction processing of the filled area is the judgment (judgment 3) as to whether or not the texture area is to be made continuous. Hereinafter, the above-mentioned three determinations will be described in detail. There are various main image features for each judgment, such as the number of black pixels, the number of pixels for each color, the density co-occurrence matrix, and the spatial frequency characteristic. Here, the spatial frequency characteristic using the Fourier power spectrum is used. The image feature amount and the number of black pixels based on the image feature amount will be described.

First, a two-dimensional Fourier power spectrum of image data obtained by reading a frequently used texture image as shown in FIGS. 20 (a), 20 (c) and 20 (e) with a scanner is shown in FIG. 20 (b), Shown by (d) and (f). Further, FIG. 21B shows a two-dimensional Fourier power spectrum of an image which is not a texture as shown in FIG. The p and q axes of the two-dimensional power spectrum are values proportional to the spatial frequency in the x and y directions, respectively. Comparing the power spectra of FIG. 20 and FIG. 21 described above, in the texture image, the spatial frequencies having the outstanding power are regularly scattered in the spatial frequency domain. On the other hand, the power spectrum of an image that is not a texture has no regularity of dominant frequencies, and the power of low-frequency components is dominant at random. The texture region is extracted in consideration of the above characteristics of the power spectrum. 20 and 21 shown above are obtained by calculating the Fourier power spectrum by assigning white data to -1 and black data to 1 for a monochrome binary image.

Judgment 1 in FIG. 18 is whether or not the image of the section of interest is a texture. In the pre-processing, regarding the number of black pixels for each section that has been counted and stored in advance,
Judgment is made based on equation (8). The coordinates of the section of interest are (i, j), and a1 and a2 are optimized threshold values. a1> pixels (i, j, code_black)> a2 (8) Next, consider the following parameter_k and parameter_l. This parameter takes into consideration the characteristics of the power spectrum of the texture image described above. parameter_k = Σkmax '(n) (9) parameter_l = Σlmax' (n) (10) where kmax '(n) and lmax' (n) are the spatial frequency kmax (n) of the nth largest power amplitude, It is expressed by the following equation using lmax (n). kmax '(n) = kmax (n); kmax (n) = <m / 2 (11) kmax' (n) = m-kmax (n); kmax (n)> m / 2 (12) lmax '( n) = lmax (n); lmax (n) = <m / 2 (13) lmax '(n) = m-lmax (n); lmax (n)> m / 2 (14) where kmax (n) , Lmax (n) are spatial frequencies in the x and y directions, respectively. This parameter parameter_k, paramet
When the value of er_l exceeds a certain threshold, it is determined that the power concentration to the predominant spatial frequency is high, and the texture image is considered. It can be expressed by equation (15). (15)
In the formula, a3 is an optimized threshold value. parameter_k> a3 or parameter_l> a3 (15) For the subsequent continuation of the texture region, the spatial frequency at the nth largest power amplitude is kmax.
As _st '(n) and lmax_st' (n), the amplitude of the nth largest power is stored as F_st (kmax_st '(n), lmax_st' (n)).

Judgment 2 in FIG. 18 is to judge whether the extracted texture image of the continuation start section is the same as the already coded texture image. Therefore, in this evaluation, it suffices to obtain the spatial frequency characteristic of each section and judge whether the predominant spatial frequency and its power value are equal. The specific formula is shown below.
The spatial frequency of the nth largest power amplitude of the texture image already coded is kmax_c '(n), lmax_
Let c '(n). In addition, the nth largest power of the texture image already coded is F_c (kmax_c '(n), lmax
If _c '(n)), the condition for judging that the target section and the texture image already coded are equal is the following expression. kmax_st '(n) = kmax_cm' (n) (16) lmax_st '(n) = lmax_cm' (n) (17) F_cm (kmax_cm '(n), lmax_cm' (n)))-c1 <F_st (kmax_st ') (n), lmax_st '(n))) <F_cm (kmax_cm' (n), lmax_cm '(n))) + c1 (18) (18) In the formula, c1 is a constant.

In the decision 2 in FIG. 18, if none of the texture images already coded can be found to be equal, the above kmax '(n) is set in the texture code data creation process. , Lmax '(n), F
The value of (kmax '(n), lmax' (n)) is stored as the feature amount of the texture image of the texture code m. At that time, as shown in FIG. 9, max_cm ′ (n), lmax_cm ′ (n), F (kmax_cm ′ (n), lmax are associated with a texture image of a predetermined size.
It is stored in the variable of _cm '(n)).

Further, the judgment 3 in FIG. 19 is a judgment as to whether the texture image of the continuation start section and the texture image of the target section are the same in the continuation processing. First, the number of black pixels counted in the pre-processing is compared with that in the continuation start section. The above equation is shown below. The coordinates of the continuation start section are (i_st, j_st),
Let the coordinates of the section of interest be (i, j). pixels (i_st, j_st, code_black) + c2> pixels (i, j, code_black)> pixels (i_st, j_st, code_black)-c2 (19) (19) where c2 is a constant. When formula (19) is satisfied, the following judgment is made. Next, the Fourier power spectrum of the section of interest is obtained, and the predominant spatial frequency having the nth largest power amplitude is kmax '(n), lmax' (n), and its power amplitude is
Compare them to those of the serialization start partition as F (kmax '(n), lmax' (n)). The equations are shown in the following equations (20), (21) and (22). kmax '(n) = kmax_st' (n) (20) lmax '(n) = lmax_st' (n) (21) F (kmax_st '(n), lmax_st' (n)))-c3 <F (kmax ' (n), lmax '(n))) <F (kmax_st' (n), lmax_st '(n))) + c3 (22) When the above equations (20), (21), (22) are satisfied , The section of interest is labeled with the same value as the same texture image of the continuous start section.

Once the extraction of the texture area in units of sections is completed, in the processing around the texture section,
Extraction processing is performed on the remaining portion and edge portion of the texture area that could not be extracted by the division unit processing for each area. In this process, the change in the image feature amount of the texture image is examined from the area of the texture section that has already been extracted to the outside, and the point where the change is large is determined to be the edge portion. At this time, as the first step, the change of the image feature amount is examined in the vertical and horizontal directions from the texture region extracted in units of sections. In FIG. 22A, the gray area indicated by k = 1 is a texture area that has already been extracted in units of sections, and the direction of scanning from the texture area is indicated by an arrow. Further, the texture area and the edge extracted in the processing of the first step are shown in FIG. Then the second
In the step, the change of the image feature amount of the texture image is similarly examined from the texture region extracted in the first step in the two directions as shown by the arrow in FIG.

The details of the above-described processing for checking the change in the image feature amount are as follows. In FIG. 22A, raster scanning is performed in units of sections, and sections not extracted as texture sections are noted. Next, if there is a texture section among the four sections above, below, left, and right of the section of interest, as shown by the arrow in FIG. Examine the changes. Here, a one-dimensional power spectrum is adopted as the image feature amount. As shown in the left end of FIG. 22 (c), the process in the case where a texture region section exists on the upper side is shown in FIG.
This will be described with reference to (d) and (e). Therefore, in FIGS. 22D and 22E, the upper section is the texture section and the lower section is the target section.
First, as shown in FIG. 22 (d), a one-dimensional window is set in a direction of obtaining an edge within a section of a texture area, and the power spectrum calculated therein is stored as a master spectrum G (P). The thick vertical lines in FIGS. 22D and 22E represent windows. At this time, the black data is set to 1 and the white data is set to -1, as in the case of the two-dimensional power spectrum calculation. Here, an example of the master spectrum is shown in FIG. Next, as shown in FIG. 22 (e), the power spectrum G ′ (p) is calculated while shifting the window by one pixel in the direction of obtaining the edge. Power spectrum when edge part is included in window as shown in Fig. 22 (d)
An example of G '(p) is shown in FIG. 23 (b). Sum of the differences between the shifted power spectrum and the master spectrum Σ (G (p) -G '
(p)) is calculated and the value is compared with the threshold value a4 according to the equation (23). It is determined that the window satisfying Σ (G (p) -G '(p))> a4 (23) (21) is approaching the edge part of the texture area. The change of Σ (G (p) -G '(p)) is shown in FIG. 23 (c). In FIG. 23 (c), the vertical axis is Σ
It is a value of (G (p) -G '(p)), and the horizontal axis represents the distance in which the window is displaced in the number of pixels. Just 22 in this example
When the window is shifted by pixels, the window is approaching the edge, but at this time, it is noticeable that the sum Σ (G (p) -G '(p)) of the difference from the master spectrum becomes considerably large. . In this way, the black pixel at the end portion of the window that is determined to approach the edge portion is set as an edge pixel, and D (x, y) = k.

In the edge extraction of the texture area,
The edge part is not always extracted as a closed loop. Therefore, closed-loop processing of edge pixels is performed.
The closed loop process is basically a continuous process of edge pixels, and when there are no more edge pixels to be continuous,
It is determined whether the condition of being a closed loop is satisfied. In this continuation process, the continuation process for each partition in the extraction of the above-described filled area is performed for each pixel. If the condition of being a closed loop is not satisfied, the black pixel around the edge pixel where the continuation process ends is newly set as an edge pixel, and the continuation process is restarted. If the condition of being a closed loop is satisfied when the continuation is completed, the edge closed loop processing is ended.

The closed loop processing will be described in detail with reference to FIG. FIG. 24 shows the image processing of the embodiment of the present invention.
It is the figure which showed intelligibly the process which makes the edge pixel of the extracted texture area into a closed loop. First, a raster scan is performed from the upper left to search for edge pixels with D (x, y)> = 1. In FIG. 24, the edge pixel is represented by a black section. Further, in the labels attached to the sections representing the edge pixels, A indicates the target pixel and a indicates the continuous edge pixels. When one edge pixel is found, the pixel is used as a starting point and the continuation process is started. In FIG. 24 (a), the label of A is set to the continuation start pixel, and the pixels that are serialized by the surrounding pixel scan from the continuation start pixel are
Label a. At this time, the continuation condition is D
This means that the (x, y) values are equal. This continuation process is continued until a branch pixel is found. As shown in FIG. 24B, the branch pixel is an edge pixel that satisfies the condition for the surrounding pixels that "the diagonally lower right is not an edge pixel, but the adjacent pixels on the right and the lower are edge pixels". In this example, it is found that the pixel of label A satisfies the condition of the branch pixel when the continuation of the pixel of label A shown in FIG. In FIG. 24 (d), the branch pixels are hatched. In addition, the x and y coordinates of the branch pixel are stored.

The continuation process after finding the branch pixel is
The label assigned to the pixel is changed according to the position of the pixel. This continuation process is performed in two stages as described below. : Pixels labeled before finding a branch pixel (pixels labeled a in Fig. 24 (c)) by peripheral scanning continuation process: After finishing the process, a pixel labeled after finding a branch pixel Continuation processing by peripheral scanning of (pixels of labels b and c in FIG. 24 ()) In the first processing, the value of the label indicating the continuous area is changed at the branch pixel in the x direction. Figure 24
(e) shows an example of the peripheral scanning result from the pixels that are labeled before the branch pixel is found. Label the left pixel including the branch pixel with c. Conversely, FIG.
As shown in 4 (f), the label on the right side of the branch pixel is labeled b. FIG. 24 (f) shows a state in which the process of FIG. Next, in the process of (1), the labels newly given to the continuous edge pixels are made equal to the target pixel. Therefore, the pixel that is made continuous from the pixel labeled b is b
, And the label of b is attached to the pixel which is made continuous from the pixel of label of c. FIG. 24 (g) shows a state in which the above processing is completed. When there are no continuous edge pixels, as shown by b and c in FIG. 24 (g),
When it is confirmed that the different labels are adjacent to each other, the closed loop formation is successful.

On the other hand, when pixels labeled with different labels are not adjacent to each other, the edge portion is not a closed area. In such a case, it is determined from the last pixel in the labeling order whether or not there is a black pixel that has not been extracted as an edge pixel around it. If such a black pixel exists, the black pixel is newly labeled as an edge pixel, and the continuation process is restarted from there. For example, in FIG. 24 (h), the pixels indicated by the dot pattern sections represent black pixels that have not been extracted as edge pixels. Also, there is an extracted edge pixel below the dot section. In such a case, the black pixel adjacent to the continuous edge pixel group is more likely to be an edge than the extracted edge pixel. I think it seems. Therefore, as shown in FIG. 24 (i), a black pixel that is not an edge pixel is newly labeled as an edge pixel, the continuation process is restarted, and this process is repeated until it finally becomes a closed loop.

Next, the color mode processing will be described in detail. As shown in the flowchart of FIG. 5, color code data creation processing first comes in, but the subsequent large processing flow is almost the same as the mono mode processing. First, the color code data creation will be described.
Data is reorganized in advance by color code. This will be described with reference to FIGS. 25 (a), 25 (b) and 25 (c). Full-color raster data is usually composed of a list of (R, G, B) as shown in FIG. In this data format, complete color information cannot be obtained unless all R, G, and B data are accessed for image processing, which requires processing time and is inefficient. Therefore, considering the variation of the data when the image is read by the scanner, as shown in FIG.
In the G, B color space, a certain area is color code 1, 2,
Defined as 3. Then, within that area (R, G,
B) The code of the color area is assigned as the data of the pixel having the data. That is, the pixel data having data such as (r1, g1, b1) is set to 1, and this is used as a color code. As described above, by dividing the color space and defining some color codes, as shown in FIG.
The G, B) data is converted into a list of color codes as shown in FIG. In the case of the color mode, this color code data is A (x, y).

Next, as in the mono mode, the number of pixels of each color in the section as shown in FIG. 7B is counted in the preprocessing. For example, the number of pixels of the color code 1 at this partition coordinate (i, j) is expressed as pixels (i, j, code_1). Since the number of pixels of each color for each section is necessary for the subsequent processing, it is stored in the memory 103.

Next, according to an algorithm described later, the filled area and its edge portion are extracted, and the area number k is assigned to the filled area in the same order as in the mono mode. Further, as shown in the equations (1) and (2), the values of the raster data B (x, y) of the filled area and the raster data D (x, y) of the edge portion of the filled area are set to the area number k. Here, the filled area is an area of one color, and the edge portion is one pixel on the outermost side of the filled area. Also,
As shown in FIG. 26, the area number of the filled area and the color code of the area are stored in association with each other. Next, the filled area and the edge portion are deleted from A (x, y). Specifically, B (x,
The value of A (x, y) of a pixel that is not y) = 0 or D (x, y) = 0 is set to 0.

Next, the texture region and its edge portion are extracted according to the algorithm described later. The target of extraction here is a texture image composed of two colors. The edge of the texture area is a closed loop that surrounds the texture area, and the edge is one of two colors that form the texture image. When one of the two colors is white, the closed loop of the other color is automatically extracted as an edge. Here, as in the mono mode, the area number k is assigned to the extracted texture area. And
As shown in equations (3) and (4), the raster data C (x, y) of the texture area, the raster data of the edge portion of the texture area
The value of E (x, y) is set to the area number k. At this time, as shown in FIG. 27, the area number of the texture area, the texture code, and the color code are stored in the memory 103 in association with each other. Since the texture code corresponds to a pattern expressed by two colors of black and white, the color texture image is expressed by the color code and the texture code by associating the smaller number of the color code with white and the larger number with black. it can. As in the case of the mono mode, this texture code is registered with both the raster data and the image feature amount of the texture image when the texture image is of a new type. As shown in FIG. 9, raster data of a texture image of a predetermined size is stored in the memory 103 in association with the code. Next, the texture area and its edge portion are deleted. Similar to the filled area, the value of A (x, y) in the texture area where C (x, y) = 0 or E (x, y) = 0 is not set to 0.

After deleting the filled area and the texture area, line image processing is next performed. This target is A
The data is (x, y), D (x, y), and E (x, y). A (x, y) is a color code, and D (x, y) and E (x, y) are data of area number k. In color mode, one vector consists of one color.
Therefore, A (x, y) is thinned and vectorized for each color code. Then, a color code is added to the final image description language as an attribute of vector data. Next, thinning is performed for D (x, y) and E (x, y), and the values of D (x, y) and E (x, y) are set to 1 in a series of line image processing as in mono mode. When the value is equal to or more than the above value, the image data is black, and when 0, the image data is white. Also in the color mode, different processing is performed in the contour mode and the whole image mode as in the mono mode. D for all image modes in color mode
After vector-converting the edge image data of (x, y) and E (x, y), the image in the closed region is extracted from the region number in FIG.
The color code of the filled area, the texture code of the texture area, and the color code shown in 27 are referenced, and these codes are added to the final image description language as attributes of vector data. In this way, A (x, y), D (x,
When vectorization is completed for y) and E (x, y), the end points and passing points of the vector are obtained as shown by the black circles in the figure of FIG. 7 (h). When vectorization is completed, it is converted into an image description language as shown in FIG. 28A shows the case of the contour mode, and FIG. 28B shows the case of the full image mode. The meaning of the codes in these image description languages is the same as in mono mode. However, the color code is written at the end of each code.

Each processing described above will be described in detail below. First, the method of extracting a filled area is basically the same as the mono-mode filled area extraction processing shown in FIG. 11, and therefore the points different from the mono mode will be mainly described below. The variables used here are the same as in the mono mode. First, there are the conditions for determining the filled area. First, the condition of the continuous start section is a color code that satisfies the following equation (24).
Whether or not n exists. In the equation, a5 is an optimized threshold value. pixels (i, j, code_n)> a5 (24) As shown in Fig. 12 (b) in mono mode, when one partition of the filled area that satisfies Eq. (24) is found, that partition (i, j
The continuation process is started from j). A flowchart of the continuation process is shown in FIG. 14 as in the mono mode. By this continuation processing, consecutive filled sections of the same color are labeled with the same value and recognized as one continuous area. This continuation process is almost the same as that in the mono mode, but the judgment for continuation is slightly different from that in the mono mode, and a condition that "the color code is the same" is added. If this is concretely expressed by an equation, it becomes as shown in equation (25).
This is the case where the color code having the number of pixels exceeding the threshold value of a5 in the continuous start section is code_m. pixels (i, j, code_m)> a6 (25)

After the extraction of the filled area in units of sections is completed, in the processing around the filled section as in the mono mode, the remaining portions and edges of the filled areas that cannot be extracted in units of sections for each area are processed. Extract the part. This processing is also almost the same as the processing in the mono mode, and the determination of the continuity of the pixels changes to "whether the pixels have the same color code".

Next, a method of extracting a texture area will be described. The texture region extraction process is the same as the mono mode process. Initialized variables C (x, y), E
It is to change the value of (x, y) to the area number. In the case of the all image mode, as shown in FIG. 27, the area number is stored in association with the texture code and the color code. The processing is also almost the same as the flowcharts shown in FIGS. Differences are judgments 1, 2, and 3 in the flowcharts of FIGS. In addition to this, when calculating the Fourier power spectrum, a pre-processing of assigning the color code data of A (x, y) to -1 and 1 is required. The rule of this pre-processing is that the smaller color code number corresponds to -1 and the larger color code number corresponds to 1.

The above three judgments will be described in detail below. Judgment 1 in FIG. 18 is whether or not the image of the section of interest is a texture. In the pre-processing, the number of pixels for each color code, which is counted and stored in advance, is first determined. It is assumed that the coordinates of the section of interest are (i, j), and the first condition is that there are color codes p, q that satisfy both of the following expressions (26) and (27). a7> pixels (i, j, code_p)> a8 (26) pixels (i, j, code_p) + pixels (i, j, code_q) ≈ number of pixels in one section (27) Next, the following parameter_k, parameter_l This parameter is equal to that of mono mode. The judgment conditions are also parameters parameter_k, paramet
Whether or not the value of er_l exceeds a certain threshold. When the above judgment condition is cleared and one texture section is found, the above n is set for the subsequent continuation of the texture section.
The spatial frequency of the second largest power amplitude
kmax_st '(n), lmax_st' (n), and the amplitude of the nth largest power is F_st (kmax_st '(n), lmax_st' (n))
Is stored as in the mono mode. In the color mode, the number of pixels of the two color codes forming the texture image is stored as pixels (code_st1) and pixels (code_st2) in ascending order of the numbers of the color codes.

Judgment 2 in FIG. 18 is to judge whether the extracted texture image of the continuation start section is the same as the texture image already coded. Therefore, this evaluation compares the value of the dominant spatial frequency, its power, with the value of the texture image already coded.
This judgment is the same as in mono mode. Figure 1
In the decision 2 in 8, if the texture image already coded is not found to be equal, in the texture code data creation processing, kmax '(n), lmax' ( n), F (kmax '(n), lmax'
(n)) and the image data of a predetermined size are newly added to kmax_c.
1 '(n), lmax_c1' (n), F (kmax_c1 '(n), lmax_c1' (n))
Remember. This process is also the same as the mono mode process.

Further, the judgment 3 in FIG. 19 is a judgment as to whether the texture image of the continuation start section and the texture image of the target section are the same in the continuation processing. First, the number of pixels of the two color codes forming the texture image, which has been counted in the pre-processing, is compared in the target section with that in the continuation start section. The formula is shown below. Set the coordinates of the continuation start section to (i_st, j_s
Let (t) and the coordinates of the section of interest be (i, j). pixels (i, j, code_st1) + c4> pixels (code_st1)> pixels (i, j, code_st1)-c4 (28) pixels (i, j, code_st2) + c4> pixels (code_st2)> pixels (i, j , code_st2) -c4 (29) When the above two expressions are satisfied, the following spatial frequency characteristics are compared. In this judgment, the Fourier power spectrum of the section of interest is obtained, the predominant spatial frequency having the nth largest power amplitude is kmax '(n), lmax' (n), and the power amplitude is F (kmax '( n), lmax '(n)) and compare them with those of the continuation starting partition. kmax '(n) = kmax_st' (n) (30) lmax '(n) = lmax_st' (n) (31) F (kmax_st '(n), lmax_st' (n)))-c5 <F (kmax ' (n), lmax '(n))) <F (kmax_st' (n), lmax_st '(n))) + c5 (32) When the above equations (18), (19) and (20) are satisfied , The section of interest is labeled with the same value as the same texture image as the continuous start section.

In general, after the extraction of the texture area in units of sections is completed, in the processing around the texture section,
Extraction processing is performed on the remaining portion and edge portion of the texture area that could not be extracted by the division unit processing for each area. This process scans the continuity of the texture image outward from the area of the texture segment that has already been extracted. This process is basically the same as the mono mode process shown in FIG. 22, but when the continuity of the texture image is checked by the one-dimensional spatial frequency characteristic while moving the window, the color code information is displayed. Add.

The process of evaluating the continuity of the texture image from the section of the texture area shown in FIGS. 22D and 22E to the section of interest will be described in detail below. As in the mono mode, as shown in FIG. 22D, a one-dimensional window is set in the direction in which the edge is obtained within the section of the texture area. The power spectrum is calculated therein and stored as the master spectrum G (p), but in the case of color, the process of data reconstruction is performed before this. As a rule at this time, the smaller number of color code data is -1
Then, set the larger one to 1. After that, the power spectrum is calculated while shifting the window, and the difference between the master spectrum and the power amplitude is calculated. At this time, it is checked whether or not there is a pixel of a color code other than the color code forming the texture image of interest in the shifted window. Since the window is sequentially shifted to the outside, pixels of different color codes should appear at the end portion of the window, and it is determined that the outside is a region different from the texture region of interest.
Therefore, pixels up to one pixel inside the pixels of different color codes are set as the texture area, and the last pixel is set as the edge pixel. If one of the two colors forming the texture image is white, the edge is a color code other than white. Therefore, instead of the black pixel in the mono mode, the edge is obtained by targeting the pixel of the color other than white. This process is almost the same as the mono mode. When the sum of the difference between the master spectrum and the power spectrum exceeds a certain threshold,
Pixels other than white near the end of the window are used as edges. Similarly to the mono mode, D (x, y) = k of the edge pixel is set. Finally, closed loop processing is performed. Since the color information is no longer relevant at this point, this processing is all the same as the mono mode processing.

[0063]

As described above, according to the present invention, the texture region and its contour (edge) are automatically extracted from the image data, and the extracted texture region is deleted from the image data. Since the contour is thinned, it is not necessary to vectorize all the image patterns in the texture area. As a result, only necessary information can be vectorized without human intervention, which is also effective in reducing the data size.

Further, according to the present invention, the texture region is vectorized as it is as a polygon having a uniform texture image, and the image description language is made to have the type of the image in the closed region as an attribute. As a texture area, the texture area can be vectorized by validating the image information in the closed area. When separately editing vector data, it becomes possible to edit the coordinates of the closed region while the texture image is being filled.

Further, according to the present invention, the filled area and its outline portion in the image data are automatically extracted, the extracted filled area is deleted from the image data, and the outline portion of the filled area is thinned. Therefore, it is not necessary to vectorize all the image patterns in the texture area, and it is not necessary to thin the filled area. As a result, only necessary information can be vectorized without human intervention, which is also effective in reducing the data size.

Further, according to the present invention, since the filled area is vectorized as a polygon whose inside is filled and the image description language is made to have the type of the image in the closed area as an attribute, the filled area is As a filled area, it is possible to vectorize the information of the image in the closed area. When separately editing the vector data, it becomes possible to edit the coordinates of the closed region in the filled state.

Further, according to the present invention, the color forming the image is coded, the image data is reconstructed using the code of this color, the image represented by the color code is vectorized, and the code of this color is Since the image description language is used as the attribute of the vector data, even in the case of a color image, it is possible to perform vectorization in which the color information is valid. Therefore, the color can be edited in the subsequent editing of the image data.

Further, according to the present invention, the color forming the image is coded, the image data is reconstructed using the code of this color, and the texture region is automatically extracted from the image represented by the color code. The color code of the extracted texture area is converted into an image description language as an attribute of the texture area. Therefore, even in the case of a color image, vectorization in which the color information is valid can be performed. Therefore, the color of the filled area can be edited.

Further, according to the present invention, the color forming the image is coded, the image data is reconstructed using the code of this color, and the filled area is automatically extracted from the image represented by the color code. Since the extracted color code of the filled area is converted into an image description language as an attribute of the filled area, it is possible to vectorize the image information and the color information in the closed area effectively.

[Brief description of drawings]

FIG. 1 is a block diagram of an image processing apparatus according to an embodiment of the present invention.

FIG. 2 is a diagram showing an image processing mode setting screen of the image processing apparatus according to the embodiment of the present invention.

FIG. 3 is a diagram illustrating an image processing mode setting screen in a state where the image processing apparatus according to the exemplary embodiment of the present invention transits to a character area setting mode.

FIG. 4 is a diagram showing image data (a) input to the image processing apparatus according to the embodiment of the present invention, character area data (b) divided by character area division, and graphic area data (c).

FIG. 5 is a flowchart showing a main routine of image processing according to the embodiment of the present invention.

FIG. 6 is a diagram showing definitions of variables and symbols used in image processing according to the embodiment of the present invention.

FIG. 7 is a diagram showing a process of processing image data in the image processing apparatus according to the embodiment of the present invention.

FIG. 8 is a diagram showing correspondence between area numbers and texture codes of texture areas extracted by image processing according to the embodiment of this invention.

FIG. 9 is a diagram showing a correspondence between a texture code of a texture region extracted by image processing of the embodiment of the present invention, raster data of a predetermined size of this texture image, and an image feature amount of this texture image.

FIG. 10 is a diagram showing the image description language finally output by the image processing according to the embodiment of the present invention separately for the mono contour mode (a) and the mono all image mode (b).

FIG. 11 is a flowchart showing a processing of extracting a filled area in the image processing of the embodiment of the present invention.

FIG. 12 is a diagram showing a process of extracting a filled area in the image processing of the embodiment of the present invention in an easy-to-understand manner.

FIG. 13 is a diagram showing an order of running around the periphery of the peripheral scan during the continuation process in the image processing of the embodiment of the present invention.

FIG. 14 is a flowchart showing a continuation process in the image processing of the embodiment of the present invention.

FIG. 15 is a diagram showing, in an easy-to-understand manner, an edge extraction process of a filled area in the image processing of the embodiment of the present invention.

FIG. 16 is a flowchart showing processing around a filled section in the image processing according to the embodiment of this invention.

FIG. 17 is a flowchart showing the intra-section continuous scanning process in FIG. 16 in the image processing according to the embodiment of the present invention.

FIG. 18 is a flowchart showing texture region extraction in the image processing of the embodiment of the present invention.

FIG. 19 is a flowchart showing a texture segment continuation process in the image processing of the embodiment of the present invention.

FIG. 20 is a diagram showing texture images (a), (c), and (e) and their Fourier power spectra (b), (d), and (f) which are the targets of image processing according to the embodiment of the present invention. .

FIG. 21 is a diagram showing an image (a) and its Fourier power spectrum (b) which are not the targets of image processing according to the embodiment of the present invention.

FIG. 22 is a diagram showing, in an easy-to-understand manner, the edge extraction processing of the texture area in the image processing of the embodiment of the present invention.

FIG. 23 is a diagram showing a change in power spectrum when a window is moved while being applied to an edge portion of a texture image which is a target of image processing according to the embodiment of the present invention.

FIG. 24 is a diagram showing, in an easy-to-understand manner, a process of forming an edge pixel of the extracted texture region into a closed loop in the image processing of the embodiment of the present invention.

FIG. 25 is a diagram showing the correspondence between R, G, B data and each color code that is referred to when creating color code data in the color mode processing in the image processing of the embodiment of the present invention.

FIG. 26 is a diagram showing the correspondence between the area number and the color code of the filled area extracted by the color mode processing in the image processing according to the embodiment of the present invention.

FIG. 27 is a diagram showing correspondences between area numbers of texture areas, texture codes, and color codes extracted by the color mode processing in the image processing according to the embodiment of the present invention.

FIG. 28 is a diagram showing the image description language finally output by the color mode processing in the image processing of the embodiment of the present invention separately in the case of the color contour mode (a) and the color full image mode (b). Is.

FIG. 29 is a block diagram of a conventional image processing apparatus.

FIG. 30 is a flowchart illustrating an operation of a conventional image processing device.

FIG. 31 is a diagram for explaining a graphic processing process of a conventional image processing apparatus.

[Explanation of symbols]

 101 image input unit 102 arithmetic unit 103 memory 104 image output unit 105 image display unit 106 pointing device 107 OCR 201 input image display area 202 character region setting button 203 contour mode button 204 all-image mode button 205 color mode button 206 mono mode button 207 Processing start button 301 Split decision button 302 Setting cancel button 303 Mode end button

[Procedure amendment]

[Submission date] December 9, 1994

[Procedure Amendment 1]

[Document name to be amended] Statement

[Name of item to be corrected] 0023

[Correction method] Change

[Correction content]

As described above, the character area is divided, the image processing mode is set, and when the operator clicks the processing start button 207, the image processing is executed according to the flow of FIG. 5, and the raster data of the image is finalized. Is converted into an image description language. First, for convenience, the definitions of variables and symbols used in this description will be described with reference to FIG. The coordinates of the section are (i, j) and the coordinates of the pixel are (x, y). The directions of the coordinate axes are as shown in FIG. In addition, the length of one side of the section is m. In addition, the bottom right section coordinate is (imax, jma
x) and the pixel coordinates are (xmax, ymax). In addition, the variables of the image data handled in this processing are listed below. Image raster data: A (x, y) Fill area raster data: B (x, y) Edge area raster data: D (x, y) Texture area raster data: C (x, y) Raster data of the edge portion of the texture area: E (x, y) A (x, y) is set to 0 when the pixel at the coordinates (x, y) is white and is set to 1 when it is black. In the color mode, an integer of 2 or more is assigned as a color code to colors other than white and black. In addition, data B (x, y), C (x, y), D (x, y), and
The values of E (x, y) are all area numbers, and are 0 for pixels that are not extracted as areas. The outline of the processing will be described below with reference to the flowchart of FIG. First, the selection of the mono mode and the color mode is determined. Hereinafter, first, the mono mode processing will be described in detail with reference to FIG. 7. FIG. 7 is a diagram showing a process of processing image data in the image processing apparatus according to the embodiment of the present invention. Figure 7
5 includes image data A (x, y) obtained by the processing of FIG. 5 and B (x, y), C (x, y), D (x, y), E (x, y) formed during the processing. It is shown. 7 (a) to (g), each data is 0
Pixels are displayed in white, and one or more pixels are displayed in black.

[Procedure Amendment 2]

[Document name to be amended] Statement

[Correction target item name] 0035

[Correction method] Change

[Correction content]

The continuity of black pixels is judged by whether or not two adjacent lines overlap each other, and by repeating this, the filled area is extracted. This will be described step by step. FIG. 15 (d) shows the outermost one line of the filled area and the adjacent one line further outside, which are the first noticed lines of this processing. First, in the inner one line, the x coordinate x_in1 of the start point and the x coordinate x_in2 of the end point of the black pixel row are obtained. Next, the x coordinate x_out1 of the start point and the x coordinate x_ou of the end point of the outer black pixel row
Find t2. If x_out2> x_in1 or x_in2> x_out1 (6) is judged in these two lines and either of the above two expressions is satisfied, the continuity is confirmed by overlapping these two lines. And
Since it is confirmed that the outer black pixel row is a filled area, B (x, y) = k (k; area number).
Next, as shown in FIG. 15 (e), in order to confirm the continuity of the line further outside by one pixel, the x-coordinates of the start point and the end point of the outer black pixel row are newly set as the coordinates of the inner black pixel row. To do. x_in1 = x_out1, x_in2 = x_out2 (7) Similarly, until a new black pixel row is not extracted to the outside, or the extracted black pixel row does not overlap with the inner black pixel row for one pixel at all, the above The process of is repeated. With the above processing, the filled area can be extracted.

[Procedure 3]

[Document name to be amended] Statement

[Name of item to be corrected] 0037

[Correction method] Change

[Correction content]

Next, a method of extracting a texture area will be described. FIG. 18 is a flowchart showing the texture region extraction processing. FIG. 18 is a flowchart of texture region extraction in image processing according to the embodiment of the present invention. In the figure, k is a region number. The process of extracting the texture region is basically the same as that for the filled region. First, the texture region is extracted in the unit of the divided sections in the pre-processing, and then the region of the pixel unit is extracted in the processing around the texture section.

[Procedure amendment 4]

[Document name to be amended] Statement

[Correction target item name] 0055

[Correction method] Change

[Correction content]

Each processing described above will be described in detail below. First, the method of extracting a filled area is basically the same as the mono-mode filled area extraction processing shown in FIG. 11, and therefore the points different from the mono mode will be mainly described below. The variables used here are the same as in the mono mode. First, there are the conditions for determining the filled area. First, the condition of the continuous start section is a color code that satisfies the following equation (24).
Whether or not n exists. In the equation, a5 is an optimized threshold value. pixels (i, j, code_n)> a5 (24) As shown in Fig. 12 (b) in mono mode, when one partition of the filled area that satisfies Eq. (24) is found, that partition (i, j
The continuation process is started from j). A flowchart of the continuation process is shown in FIG. 14 as in the mono mode. By this continuation processing, consecutive filled sections of the same color are labeled with the same value and recognized as one continuous area. This continuation process is almost the same as that in the mono mode, but the judgment for continuation is slightly different from that in the mono mode, and a condition that "the color code is the same" is added. If this is concretely expressed by an equation, it becomes as shown in equation (25).
This is the case where the color code having the number of pixels exceeding the threshold value of a6 in the continuation start section is code_m. pixels (i, j, code_m)> a6 (25)

─────────────────────────────────────────────────── ─── Continuation of the front page (51) Int.Cl. ⁶ Identification code Internal reference number FI technical display location 7459-5L G06F 15/70 335 Z

Claims (7)

[Claims] 1. A means for automatically extracting a texture region such as hatching (a region having a uniform change in brightness and color is referred to as a texture region) and its contour (edge) in an image processing apparatus that handles raster data. An image processing apparatus comprising: a means for deleting the extracted texture area from the image data; and means for thinning the contour of the texture area. 2. The image processing apparatus according to claim 1,
Further, the image processing is characterized by having means for vectorizing the contour portion of the thinned texture area as a polygon and means for making an image description language so that the type of the image in the closed area has an attribute. apparatus. 3. In an image processing apparatus handling raster data, means for automatically extracting a filled area and its contour portion in the image data, and the extracted filled area being deleted from the image data, and a contour of the filled area. An image processing apparatus comprising: a thinning unit. 4. The image processing device according to claim 3,
Further, the image processing is characterized by having a means for vectorizing the contour part of the thinned filled area as a polygon, and a means for making an image description language so that the type of the image in the closed area has an attribute. apparatus. 5. An image processing apparatus for handling raster data, means for encoding colors forming an image, means for reconstructing image data using the color codes, and an image represented by a color code. An image processing device comprising vectorization means and means for converting the color code into an image description language as an attribute of vector data. 6. An image processing apparatus which handles raster data, means for encoding colors forming an image, means for reconstructing image data using the color code, and texture from an image represented by a color code. An image processing apparatus comprising means for automatically extracting a region and means for converting the extracted color code of the texture region into an image description language as an attribute of the texture region. 7. An image processing apparatus for handling raster data, means for encoding colors forming an image, means for reconstructing image data using the color codes, and filling from an image represented by a color code. An image processing apparatus comprising means for automatically extracting an area and means for converting the extracted color code of the filled area into an image description language as an attribute of the filled area.