JPH01295376A - Linear approximation processing system for line pattern - Google Patents

Abstract

PURPOSE:To freely set a parameter in case a line pattern is evaluated by its curvature and displacement for execution of the linear approximation by changing the displacement error value in response to the linear segment length and deciding the presence or absence of a feature point based on the result of comparison between said displacement and error value. CONSTITUTION:When the linear approximation is applied to a line pattern, the comparison is carried out between the displacement (d) of the original data from an approximate straight line and a parameter (Thd) as well as the inter- feature point distance (l) and a parameter (Thl) for decision of the presence or absence of a feature point. In other words, an evaluation parameter required for the visually preferable linear approximation is set by changing the error value (Thd) of the displacement (d) in accordance with the linear segment length (l). Then the presence or absence of a feature point is decided by comparing the displacement (d) with the value (Thd). Thus a parameter having a large freedom degree can be set in case a line pattern obtained from extraction of contours in a picture processing job is evaluated by the curvature and the displacement of said pattern for execution of the linear approximation.

Description

[Detailed Description of the Invention] [Technical Field] The present invention relates to a straight line of a line figure obtained by contour extraction, which is a type of image processing, which is approximated by a polygonal line by evaluating the line figure in terms of curvature and displacement. Concerning approximation processing methods.

[Prior art]

Conventionally, as a shape approximation method for line figures obtained by image contour extraction, etc., there has been a A line segment approximation method that preserves features is known.

In the above straight line approximation method, the rough parts are roughly approximated and the fine parts are approximated finely, and the criteria for approximation evaluation are basically the curvature approximation R (=d/f) and the displacement D (=d). It is disclosed that

However, if the above-mentioned curvature approximation R is judged using a fixed threshold value, in cases where it is not desired to make the conspicuousness of the corner simply proportional to the size, for example, Pl in Fig. 2(a) and the same figure ( b
) If you want to make a difference in the conspicuousness of the corners with P2 in the middle, this cannot be done.

The following describes the conventional method of linear approximation according to the complexity of a figure, that is, the method of linearly approximating rough parts coarsely and finely making fine parts.

The algorithm for linear approximation in this case consists of three steps, and each step will be explained below in order.

(1) Step 1 Find the bending point, starting point, and ending point when sequentially connecting the point sequence (P,) consisting of points on the grid shown in Fig. 3 (○ marks in the figure).

Note that the direction of the line segment formed by sequentially connecting the determined point sequence (Q,) is one of the eight directions shown in the figure.

(2) Step 2 From the point sequence on the grid obtained in step 1 above, find a pattern that appears when a straight line is digitized, fit the straight line, and compress the point sequence.

FIG. 4 shows an example of a point sequence pattern that can be regarded as a straight line, and a cough point sequence is considered to be on a straight line when it satisfies the following conditions.

Iso 1<2 ORIsI 1<Madashi, (
SO+Sz) or (Is,)l≧2△ IS, 1≧2
), only P,,P,02 points are considered to be on the straight line.

(3) Step 3 From the results of step 2 above, find the final approximate straight line. The criteria for evaluation at this time are the error calculation elements shown in Figure 5, namely the approximation R of curvature (=d/f) and the displacement D.
(=d) is used as a basis.

In addition, to find the maximum neighbor straight line PtPt= with P as the starting point, (Rj≦T, +)△ for all points Pj between them.
It is sufficient to find Pk for which (Ω, ≦′b) (T+, b: L threshold) holds and does not hold for the line segment PiPk*1.

The method of using the curvature approximation R and displacement as the criteria for evaluating the linear approximation described above can also be applied to the divided linear approximation to the original data.

In the following, an iterative endpoint fitting method (piecewise linear approximation method) will be explained with reference to FIG. 6 as an example. Note that the ○ marks in the figure indicate each point in the series of points.

(1) Step 1 First, as shown in (a) of the same figure, a point sequence (point segment)
Virtually connect the points at both ends of with a straight line segment, find the point (dividing point) farthest from the straight line segment, and if the maximum distance d (do in the figure) is greater than the error value Thd, or the above hypothetical Straight line segment length! The ratio d/2 (d, /12° in the case shown) of the maximum distance d (20 in the figure) is the error value Th
If it is larger than d, it is divided into two straight line segments at the above dividing point as shown in FIG.

Note that if it is neither of the above, the algorithm in this process is terminated. '(2) Step 2 Next, the same procedure as in Step 1 above is repeated for each divided straight line segment, and the straight lines shown in (C) and (d) are obtained. Divide into line segments.

(3) Step 3 For any straight line segment divided by the above algorithm, the division operation is stopped when the maximum distance d described above becomes smaller than the threshold value Th, and each division point is The algorithm ends as a line figure to find straight lines connected by .

Note that in the above-mentioned algorithm, in the case of a closed loop, l=0, so it is necessary to take measures such as dividing at least once, for example.

However, the conditional area of the feature points in the above-mentioned linear approximation method is the shaded area shown in FIG. There is a problem in that this cannot be done when it is desired to make the conditional area of the boundary point as shown in the shaded area shown in FIG. 2(b).

〔the purpose〕

The present invention was made in order to solve the above-mentioned conventional problems, and when a line figure obtained by contour extraction etc. in image processing is evaluated in terms of curvature and displacement and linear approximation is performed, the degree of freedom is reduced. It is an object of the present invention to provide a linear approximation processing method for line figures that allows setting of large parameters.

〔composition〕

In order to achieve the above object, the present invention, when linearly approximating a line figure, determines whether or not it is a feature point by comparing the displacement d of the original data with respect to the approximate straight line and the parameter Thd,
This is characterized in that it is performed based on a comparison between the distance 2 between feature points and the parameter Thf.

Hereinafter, the present invention will be specifically explained based on examples.

The evaluation parameters for the visually preferable linear approximation in the present invention are set by changing the error value Thd of the displacement d in accordance with the straight line segment length l shown in FIG. 6 described above.

To describe this using the formula, if 1=tx then Thd=β (1
). Then, by comparing the above-mentioned displacement d and the error value Thd, it is determined whether the point is a feature point or not.

A specific implementation method in this case will be described below.

For example, if an input image with a pixel count of 512 x 512 is used and a contour is extracted based on tracking processing in four directions for the input image, there are at most 1024 types of integers that the straight line segment length l can take. be. Based on this,
A look-up table (LUT) is created that outputs the error value Thd for the input of the straight line segment length 2 described above.

Further, if the relationship between the error value Thff1 with respect to the straight line segment length l and the error value Thd with respect to the displacement d is a simple function, the above error value Thd is calculated by calculation based on the following formula.

For example, if you want the above error values Thf and Thd to have a proportional relationship, Thd=kX/! (k is a certain constant) (2).

FIG. 1 is a flowchart of the linear approximation processing method according to the present invention, and each step will be explained below.

First, in step S, after contour tracing (X,
Y) Based on the point sequence data described in a coordinate sequence, it is determined whether the cough point sequence data is a closed loop.

If it is determined in step S1 that the loop is not closed, the process moves to step S, and if it is determined that the loop is closed, the process moves to step S2.

If the loop is closed, in step S2,
The above point sequence data is divided at the point farthest from the starting point (end point). Note that the division process in step S2 is based on the straight line segment length in the case of a closed loop! This is a process for performing division at least once based on the fact that 0 becomes 0.

Next, in step S3, the error value Th is calculated from the lookup table (LUT) described above for each segment.
While determining d, in step S4,
Find the maximum distance d for each segment.

Subsequently, in step S, the magnitude relationship between the maximum distance d obtained in step S and S4 described above for each segment and the error value Thd is determined. If it is determined in step Ss that the maximum distance d is greater than the error value Thd, the segment of interest is divided in step S, and then the process returns to step S described above.

Further, in step S, the maximum distance d is the error value Th
If it is determined that the segment size is smaller than d, the process moves to step S without dividing the segment of interest.

Then, in step S7, it is determined whether the point sequence data is finished or not, and if it is still in progress, the process returns to step S1 and restarts the processing in each step described above. Further, when it is determined that the processing for all point sequence data has been completed, the process moves to step S8, and in step S, the dividing points are connected, thereby completing the series of linear approximation processing.

〔effect〕

According to the present invention described above, the error value Thd of the displacement d is changed corresponding to the straight line segment length 2, and the error value Thd of the displacement d and the error value Th
Since it is decided whether it is a feature point or not based on the comparison with Linear approximation results can be obtained.

[Brief explanation of the drawing]

Fig. 1 is a flowchart showing the processing procedure in the linear approximation processing method according to the present invention, Fig. 2 is a diagram for explaining disadvantages in the conventional linear approximation method, and Fig. 3 is a grid point sequence in the conventional linear approximation method. Figure 4 is a diagram showing an example of a point sequence pattern that can be regarded as a straight line in the conventional linear approximation method, Figure 5 is a diagram showing error calculation elements in the conventional linear approximation method, and Figure 6 is a diagram showing the error calculation elements in the conventional linear approximation method. FIG. 7 is a diagram for explaining polygonal line approximation using the conventional iterative endpoint fitting method. FIG. 7 is a diagram showing the conditional range of feature points in the conventional linear approximation method. Figure 1 (G) (b) Pl 0α~Q
J Figure 3 (a) (C) (b)
(d) Figure 6 (G) (b) Figure 7

Claims (4)

[Claims] (1) In the linear approximation process in which a line figure is approximated by a straight line by evaluating it using curvature and displacement, the determination of whether or not it is a feature point is based on the displacement (d) of the original data with respect to the approximate straight line and the error value for this displacement. A certain parameter (T
hd) and a distance (1) between feature points and a parameter (Thl) that is an error value for the distance. (2) The above parameter (Thd) and the parameter (Th
1) A linear approximation processing method for a line figure according to claim 1, characterized in that a specific relationship is established between the line figure and the line shape. (3) The linear approximation processing method for line figures according to claim (2), characterized in that when the parameter (Thd) becomes large, the parameter (Thl) has a large relationship. (4) The above parameter (Thd) and the parameter (Th
Claim (3) characterized in that l) has a proportional relationship.
Linear approximation processing method for the described line figure.