JPH0495180A - Segment/curve recognizing system - Google Patents

Abstract

PURPOSE:To dispense with an image processing occupying much processing time and to exactly execute the segment/curve recognition in a short processing time by converting binary digital image data into a hierarchical data structure, and checking this data. CONSTITUTION:The system provided with a picture element connection data generating means 103 for scanning a binary digital image in the main scanning direction, and generating picture element connection data by setting a picture element varied to a picture element of an image from a background picture element as a picture element of a start point, and setting a picture element varied to the background picture element from the picture element of the image as a picture element of an end point, and a picture element block data generating means 105 for checking a mutual relation of picture element connection data, and generating an assembly of the picture element connection data having a prescribed relation as picture element block data. In such a state, the image is expressed by the picture element connection data and the picture element block data being an assembly of the picture element connection data, and based on these data, a segment and a curve are discriminated, respectively. In such a way, the linear segment and the curve in the image data can be recognized exactly and at a high speed without executing the image processing.

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to a line segment/curve recognition method that reads line graphic images as binary digital data and recognizes and extracts line segments and curves from the binary digital data.

[Conventional technology]

Conventional techniques for identifying line segments and curves include segmenting a line figure into short lines, representing them as sheet vectors, checking the change in angle between adjacent sheet vectors, and detecting when the change exceeds a predetermined value. There is a method of determining it as a curve, and determining it as a line segment when it is within a predetermined value (Japanese Patent Laid-Open No. 1-116772). In addition, as another conventional technique, a curve is fitted to the polygonal line data, the curve is approximated, and the curve is expressed.
-) There is a method of calculating the area of the part surrounded by the vector and deciding whether or not to adopt the above curve based on this area (Japanese Patent Application Laid-Open No. 62-251988).

[Problem to be solved by the invention]

However, in the case of the former method, the feature points (intersections, bending points) that make up the line figure are extracted in advance, and it is recognized whether the specific section is a straight line or a curved line. It was not possible to recognize straight lines or sections where straight lines and curves were mixed. In addition, in the case of the latter method, in order to determine whether a curve is a straight line or a straight line, a long processing time is required to fit the curve to all the seat vectors in the 1 m image. . The present invention solves the problems of the prior art described above and realizes a line segment/curve recognition method that can accurately and quickly recognize straight line segments and curves in image data without performing image processing. The purpose is to

[Means to solve the problem]

The line segment/curve recognition method of the present invention scans a binary digital image in the main scanning direction, sets a pixel that has changed from a background pixel to an image pixel as a starting point pixel, and A pixel connected data generation means (103 in FIG. 1) that generates pixel connected data with a pixel as the end point pixel, and a pixel connected data generation means (103 in FIG. 1) that examines the mutual relationship of the pixel connected data and creates a pixel block of a collection of pixel connected data having a predetermined relationship. Pixel block data generation means for generating data (105 in FIG. 1)
and a line segment/one curve recognition means that examines the generated pixel block data and identifies whether the pixel block is a straight line or a curve (Fig. 1 1.07, l O8,109,1
10).

[Effect]

In order to achieve the above object, the present invention converts an image into a hierarchical data structure and uses that data structure to
Performs curve recognition processing. That is, an image is expressed by pixel connection data and pixel block data which is a collection of pixel connection data, and line segments and curves are respectively identified based on these data. The pixel connection data generation means generates pixel connection data. Pixel connection data is generated by scanning a binary digital image in the main scanning direction, using a pixel that changes from a background pixel (white pixel) to an image pixel (black pixel) as the starting point pixel, and converting the image pixel to the background pixel. The changed pixel is set as the end point pixel, and is so-called image run information. The pixel block data generation means generates pixel block data. Pixel block data refers to a collection of pixel connected data having a predetermined relationship. The predetermined relationship is, for example: 1. Two or more pixel connected data are not connected to one pixel connected data; 2. The length of pixel connected data connected under one pixel connected data is the same as the length of the previous pixel connected data. There is a method that satisfies the following two conditions: does not change significantly with respect to the length of the pixel connected data. The line segment/curve recognition means, based on each pixel block data,
Identify whether a pixel block represents a line segment or a curve. As a specific example of the identification, for example,
■As in the embodiment of the present invention described later, block data is divided into sub-blocks each containing a predetermined number of pixel-connected data, and the top and bottom of each sub-block are A method that determines the midpoint between the start point and end point of pixel connected data and uses it as vector data, and identifies changes in the direction of the vector in each pixel block data to identify straight curves. ■ Each pixel block data There are methods to identify the linearity of pixel block data by performing Hough transform on a specific point (starting point or midpoint) of pixel connected data, and methods to identify the linearity of pixel block data by pattern matching. . According to the present invention, image data is converted into a hierarchical data structure consisting of pixel connection data and pixel block data, and line segment/curve recognition processing is performed using this data structure. When performing curve recognition processing, there is no need to perform image processing that takes processing time, and accurate processing can be performed in a short processing time.

【Example】

FIG. 1 shows the configuration of an embodiment of the present invention. The line segment/curve recognition method of this embodiment includes an image input section 101, an image data storage section 102, a pixel connection data generation section 103,
Pixel connected data storage section 104, pixel block data generation section 105, pixel block data storage section lO6, sub-block generation section 107, sub-block storage' section 108,
unit vector calculation section 109, line segment/curve recognition section 110,
Pixel block decomposition unit 111, line segment data storage unit 112,
It consists of a curve data storage section 113. An image input unit 101 inputs two images such as an image scanner.
It is input as a value. The image data storage unit 102 stores image data input from the image input unit 101. The pixel connected data generation unit 103
The binary digital image stored in 2 is scanned in the main scanning direction, and the pixel that changes from a background pixel (white pixel) to an image pixel (black pixel) is set as the starting point pixel, and the pixel changes from a black pixel to a white pixel. This is to generate pixel connection data that uses the pixel as the end point pixel. The pixel connection data storage section 104 stores data generated by the pixel connection data generation section 103. The pixel block data generation unit 105 examines pixel connected data adjacent in the vertical direction (sub-scanning direction) and generates a collection of pixel connected data having a predetermined relationship as pixel block data. The pixel block data storage section 106 stores the pixel block data generated by the pixel block data generation section 105. The sub-block generation section 107 decomposes pixel block data into sub-blocks having a predetermined brightness and stores them in the sub-block storage section 108. The unit vector calculation unit 109 calculates unit vectors representing all sub-blocks in a certain pixel block data. The line segment/curve recognition unit 110 includes a unit vector calculation unit 109
It examines changes in all unit vectors calculated in , and finds multiple line segments/line elements (line segments and curves). The pixel block decomposition unit 111 decomposes pixel block data including a plurality of line segments/line elements into pixel blocks representing a single line segment/line element, and stores the pixel block data representing the line segments in the line segment data storage unit 112. Second, pixel block data representing a curve is stored in the curve data storage section 113. Hereinafter, the procedure of the broken line processing will be explained in order with reference to FIG. Binary digital image data input by the image input unit 101 is stored in the image data storage unit 102. The pixel connected data generation unit 103 extracts pixel connected data, which is image run information, and pixel block data, which represents vertically connected pixel connected data in the image, from the contents of the image data storage unit 102. The extraction procedure of pixel connected data will be explained using FIG. Note that FIG. 4 is a diagram explaining pixel connection data. The image is scanned in the rask direction (main scanning direction), and the pixel 1s (i, j) changes from the background pixel value of the image to the line figure pixel value.
(Steps 201 to 204 in Figure 2). That is, first set the image width m1, image height n1, and the black pixel value px (step 201), and set the rask scan start position as r.
=0+J=O (step 202). It is checked whether the pixel 1s(i,,+) at the scanning position has a black pixel value px (step 203), and if it is a black pixel,
Pixel l5(i,j) is registered as a starting point (step 204). Further, scanning is continued to find a pixel Ie (+, , +)≠px whose value changes from a line figure pixel value to a background pixel value (steps 205 to 207). That is, it is determined whether the scanning has reached the end of one scanning line based on i<m (step 205), and if it has not reached the end, it is determined whether the line figure pixel value has changed to the background pixel value (step 205). 206), and if there is no change, the scanning is advanced by 1m votes (step 207), and the same process is repeated. When there is a pixel change in the determination at step 206, that is, Ie
When (i, j)≠px, the pixel 1e (i, j) is registered as the end point of the pixel connected data (step 20
8). At this time, if a pixel that changes from the line figure pixel value to the background pixel value cannot be found to the right end of the rB image, it is determined that the pixel connection has been completed on that scanning line, and the line figure pixel value changes to the background pixel value. The pixel that changes to the pixel value is Ie(m,
J). Note that pixel registration is performed in the pixel connected data storage unit 1 by setting the pixel Is as the starting point 5 (t) of the t-th pixel connected data, and setting the pixel 1e as the end point E (t) of the t-th pixel connected data.
This is done by storing it in 04. At this time, the length L (t) of the pixel connected data may also be registered as additional information. Further, this process continues until the scanning line exceeds the shoulder height (n) of the image (steps 211-
213). Next, pixel block data is extracted from the pixel connected data and image extracted above. FIG. 3 shows a flowchart regarding pixel block data extraction processing. First, set the total number of pixel connected data to m. Step 3
01), the pixel connected data number n and the pixel block data number j are set to O (step 302). The n-th pixel connected data data(n) is extracted from the pixel connected data storage unit 104, and it is checked whether this data has already been registered as part of a block (step 303). If it has already been registered, n=n+1 is set to retrieve the next pixel connected data (step 304). It is determined whether the pixel block generation process has been completed for all the pixel connected data by determining whether the pixel block is n m (step 305). If it is determined in step 303 that it has not been registered yet, new pixel block data generation processing is performed (steps 306 to 314). The process of generating new pixel block data will be explained in detail. When pixel connected data data(n) has not been registered in a pixel block, it is registered in pixel block data block(i), which is a new pixel block (step 306). Next, the value of variable n is temporarily saved for later use (step 307). Then, it is checked whether this pixel connected data data(n) satisfies the following two conditions (step 308). ■ There is only one piece of connected pixel data that is connected downward. ■ The length of the connected pixel data connected downward does not change suddenly. For example, it cannot be more than twice its own length or less than 1/2 its own length. If the pixel connected data data (n) does not satisfy the condition (■■), it is assumed that the pixel block data block (n) has ended. When the pixel connected data data(n) satisfies the condition ■■, the pixel connected data data(j) connected to this data(n) is searched from the pixel connected data storage unit 104 (step 309). Next, it is checked whether the pixel connected data data(j) satisfies the following condition (2) (step 310). (2) The connected pixel data that continues above is only data(n). When the pixel connected data data (j) satisfies the condition ■, this pixel connected data data (j) is additionally registered in the pixel block data block (i) (step 3
11). Then, the newly added and registered pixel connected data d
In order to search for pixel connection data connected to ata(j), set variable n to j (step 312), step 308
Return to The above processing is performed on the pixel block data bl. Continue until ck(i) is completed. This termination is performed in step 308 or step 310 under the conditions ■ to ■ or ■
It is determined that the condition is not satisfied, and step 304
Return to At this time, the value saved in step 307 is set to variable n (step 313), and the value of variable i is incremented (step 314). Then, the pixel connected data storage unit 104 is searched for pixel connected data that is not registered in the pixel block data, and the above processing is continued until all the pixel connected data are registered in the pixel block data. In this example, the pixel block data was explained as actually having all the pixel connected data within it, but the actual pixel block data is represented by the circumscribed rectangle of the area that the block occupies in the image. The pixel connected data included in the block is stored in the pixel connected data storage unit 10.
Only the position at 4 is recorded. An example of the result of representing an image using pixel block data is shown in FIG. Next, the sub-block generation unit 107 divides all the pixel block data stored in the pixel block data storage unit 106 into a predetermined height (for example, N pieces of pixel connected data).
Decompose into sub-blocks with . Extract N pieces of pixel connected data at the top of the pixel block data as one sub-block, and continue until the last pixel connected data of the pixel block data is reached.The last sub-block may not be able to obtain N pieces of pixel connected data, but the remaining pixel connected data can be used to extract the last sub-sub.
Block. All sub-numbered blocks extracted by the sub-block generation unit 107 as described above are stored in the sub-block storage unit 1.
Stored in 08. The stored sub-blocks must be appended with data indicating which pixel block data they belong to in order to simplify the next processing. Therefore, each sub-block is Information on where the pixel block data stored in the pixel block data is located in the pixel block data storage unit 106 is also added at the time of extraction. Next, in the unit vector calculation unit 109, each sub-block in the sub-block storage unit 108 Calculate the unit vector that represents the unit vector.However, here, only the X component of the unit vector is calculated.Figure 6 shows a diagram explaining the unit vector.The method of calculating the unit vector will be explained. Select one sub-block from section 108, extract the midpoint of the top and bottom pixel connected data of this sub-block, and use it as vector data representing the sub-block. Sawo IAI,
Assuming that the X coordinate of the starting point of A is Asx5 and the ending point is Aex, the X component λ of the unit vector of vector A is obtained by the following equation. λ= (Aex - Asx) / l A l - (Formula 1) Next, the recognition processing in the line segment/curve recognition unit 110 will be described. When the value λ calculated by the unit vector calculation unit 109 is expressed as a graph for certain pixel block data, a characteristic graph as shown in FIG. 8 can be obtained. In other words, in the case of pixel block data consisting of a single line segment (Fig. 8 (a)), the graph of λ has a slope of almost 0, and in the case of pixel block data representing a curve (Fig. 8 (C)
)), the graph of λ will either increase or decrease. Furthermore, it can be seen that in the case of pixel block data having a plurality of line segment elements (FIG. 8(b)), it can be divided into graphs with a slope of 0 by the number of line segment elements. Therefore, when the graph of λ can be decomposed into several graphs of slope O, the pixel block decomposition unit 111 divides the pixel block data at the points indicated by the graphs. In the case of pixel block data representing a curve (Figure 8 (C))
, the graph of λ increases or decreases.
This can be explained by the following. As shown in FIG. 6, the value of λ represents the value of COS ρ, where ρ is the angle formed by the X axis and the vector A. When a circle is decomposed into pixel block data, even in the worst case, it is divided into four parts as shown in Figure 7, and the vector representing the sub-block within the pixel block data expressing the left and right curves and the X axis The angle ρ made is
Since 0≦ρ≦■ (■≦ρ≦21), a graph of λ for pixel block data representing a curve always appears with an increasing or decreasing slope. The line/curve recognition unit 110 examines the graph of λ for the pixel block data as described above, and if the graph has a slope of O (if it can be divided into a few parts, the pixel block decomposition unit 111 divides the graph into small parts). The data of each divided pixel block is stored in the line segment data storage unit 112, and if the graph represents a graph with a single slope O, the data of each divided pixel block is stored in the line segment data storage unit 112 as is.
If the graph represents an increasing or decreasing slope, it is stored in the curve data storage unit 113. In this embodiment, although not specifically stated, there is pixel block data that cannot be determined as a curve or a line segment. In order to store such pixel block data, it is also possible to provide an image data storage section in the same way as the line segment data storage section and the curve data storage section. Furthermore, although this embodiment does not describe pixel block data in which curves and line segments are mixed, even for such pixel block data, the graph of λ has a slope of O and an increasing or decreasing slope. Since it has parts, by dividing it into each part, pixel block data representing a line segment and pixel block data representing a curve can be obtained. In the line segment/curve recognition unit 110, even if the slope of the graph of λ is O, the pixel block data representing the upper and lower parts of a circle or the arrow part as shown in FIG. 7 cannot be expressed as a line segment. . To avoid this,
Before storing the data in the line segment data storage unit 112, it is determined whether the data represents a line segment. Specifically, check the length of pixel connected data within the pixel block data,
If a large amount of pixel connected data of the same length exists, that pixel block data is determined to be line segment data and is stored in the line segment data storage unit 112. Further, in the case of the circle shown in FIG. 7, only the left and right portions are determined to be pixel block data representing a curve, but the remaining portions are left without being determined as either a line segment or a curve, making it impossible to recognize a complete circle. However, when examining pixel block data adjacent to pixel block data representing a curve, if the pixel block data is determined to be block data representing a curve, or is neither a line segment nor a curve, by integrating these pixel block data. , a complete circle can be expressed as one piece of data. By converting binary digital image data into a hierarchical data structure and examining changes in the direction of unit vectors within this data, it is possible to recognize whether the figure represented by each data is a line segment or a curve. thing. In addition, in this embodiment, recognition of line segments and curves is performed by dividing pixel block data into sub-blocks and using unit vectors thereof, but in addition to this, Hough transform is used to identify straight lines and curves. There is a method of identification using pattern matching of pixel blocks. For example, when using Hough transform, a specific point (starting point or midpoint) of each pixel connected data in a pixel block is calculated using the Hough transform formula p=x・
Converted to p-θ space by cosθ+yesinθ. At this time, p is the origin and the point on the pixel connection data (x, y
), θ is the distance between the origin and the point on the pixel connection data (x, y
) is the angle between the line segment connecting them and the X coordinate axis. This p-θ
From the spatial characteristics, it can be determined whether the pixel block is a straight line segment or has a plurality of straight line segments. Further, by using a Hough transform formula corresponding to the curve, it is possible to similarly identify the curve.

【Effect of the invention】

According to the present invention, by converting binary digital image data into a hierarchical data structure and examining this data,
Since the figure represented by each data is recognized as a line segment or a curve, there is no need to perform time-consuming image processing when performing line segment 0 curve recognition processing, resulting in a shorter processing time. Accurate line segment/curve recognition can be performed in a short amount of time. Furthermore, according to the present invention, since pixel block data, which has a certain size, is used, it is possible to accurately recognize line segments and curves without being affected by local variations in the figure. can. Furthermore, even if a recognition error occurs, since the pixel block data includes pixel connection data, the shape within each block can be easily known, and feedback can be easily applied to the recognition process.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the configuration of an embodiment of the present invention. FIG. 2 is a flowchart of the processing of the pixel connected data generation section of the embodiment. FIG. 3 is a flowchart of the processing of the pixel block generator of the embodiment. FIG. 4 is a diagram explaining pixel connection data. FIG. 5 is a diagram explaining pixel block data. FIG. 6 is a diagram explaining a unit vector. FIG. 7 is a diagram illustrating pixel block data for a curve. FIG. 8 is a graph showing changes in unit vectors within one pixel block data. 101... Image input section, 102... Image data storage section, 103... Pixel connection data generation section, 104...
Pixel connected data storage unit, 105... Pixel block data generation unit, 106... Pixel block data storage unit, 1
07...Sub-write block generation unit, 108...Sub-write block generation unit
Block storage unit, 109... Unit vector calculation unit, 1
10... Line segment/curve recognition section, 111... Pixel block decomposition section, 112... Line segment data storage section, 113...
・Curve data storage section. Patent applicant: Fuji Xerox Co., Ltd. Representative: Patent attorney: 2 people including Iwatori

Claims (1)

[Claims] A binary digital image is scanned in the main scanning direction, and a pixel that changes from a background pixel to an image pixel is set as a starting point pixel, and a pixel that changes from an image pixel to a background pixel is set as an end point pixel. pixel connected data generating means for generating pixel connected data to be used as pixels; pixel block data generating means for examining the mutual relationship of the pixel connected data and generating a collection of pixel connected data having a predetermined relationship as pixel block data; A line segment/curve recognition method comprising line segment/curve recognition means for examining generated pixel block data and identifying whether the pixel block is a straight line or a curve.