JPS619778A - Segment extracting system - Google Patents

Abstract

PURPOSE:To improve the reliability of an extracted segment by applying the statistic processing not only to the information obtained from a thin area on a line but to the information obtained from a broad area to extract the segment to a differential picture image. CONSTITUTION:A memory is first provided to store each of the 0th-2nd order moments to each minor area, the maximum and minimum values of the (x) coordinates, the maximum and minimum values of the (y) coordinates and the maximum and minimum values of the sloping direction of a density slope. Then the picture elements are scanned from the highest line to the lowest line as well as from the left end to the right end with an original picture image. Thus the size and the direction of the density slope are obtained with each picture element. When the size of the density slope exceeds a certain threshold level, a specific minor area to which the corresponding picture element belongs is decided in response to a range of angle difference between the direction of the slope of said picture element and the direction of the slope of each picture element in a minor area to which the picture element right above or at the exactly left of the picture element.

Description

DETAILED DESCRIPTION OF THE INVENTION The present invention relates to a line segment extraction method for extracting line segments of the outline of a target object, etc. from a gray scale image inputted by a TV left camera.

In the conventional general line segment extraction method, the original image is subjected to differential processing with no directionality in principle, and the resulting differential image is subjected to a certain threshold processing to extract the outline of the contour line as a binary image. This is done by extracting a candidate area and applying thinning processing to it.

However, this method has an ambiguous line thinning algorithm, and is vulnerable to noise because line segments are extracted by accumulating local processing as a whole. Additionally, even if it is known in advance that the final line segment is part of a straight line, the line segment is extracted without using that information, so the processing to straighten the obtained result is easy. However, there were drawbacks such as the processing time required.

The present invention has been made in order to eliminate the above-mentioned drawbacks of the conventional technology, and is capable of stably extracting line segments without being affected by noise, and without performing complicated processing. The purpose of this invention is to provide a line segment extraction method that can be performed in a short time with a small amount of processing.

Embodiments of the present invention will be described below with reference to the drawings.

In this example, first, the original image of the grayscale image shown in FIG. Find direction. This is illustrated in FIG. 2, where the direction of the arrow represents the direction of the concentration gradient, and the length of the arrow represents the magnitude of the concentration gradient. Next, a region where the concentration gradient exceeds a certain threshold is extracted to obtain the region -A (shaded region) in FIG. Next, consider this part A as one small region, which is a connected region that does not exceed the range of angular difference in which the direction of the concentration gradient differs, and A1 to A in FIG. 4.
Divide like lt. Then, each small area Al-A
t 7 whose area is smaller than a certain value, for example A1. A21As +A6 +..., etc. are excluded from processing. For each of the remaining small regions, the 0th, 1st, and 2nd moments of the concentration gradient are determined. The calculation of each moment is shown below.

0th moment "oo□=G device (X, y) l A1 1st moment mto, i-ΣX・ga (X, y) mox, i =, F, y 俤ga (x + y) 2nd moment m2o, i=X, x Φga(x+y)m□□, 1=Σ
:X”y”ga(Xly)m02. i = position 'I'
g a (X+ V )I Here, Ai i'' indicates the first small area, and g
a(x + y) represents the magnitude of the gradient at U pixel (X, y).

Next, based on these zero-order, first-order, and second-order moments, the long axis and short axis of the inertial equivalent ellipse of each small area can be calculated using the following formula. For example, as shown in FIG. 5, the inertial equivalent ellipse is a circular area d shown by a broken line with respect to the small area AI.

However, M8. , l Ml, i is It is.

The lengths of the major and minor axes of the inertial equivalent ellipse for each small region were determined as described above, and the small regions where the short axis is small and the long axis is large are line segments. It is determined that it represents
Calculate the position of the line segment.

For example, when considering the long axis of the inertial equivalent ellipse as its line segment,
Xt'l coordinates of the center of gravity (XLf
The slope θ1 centered at lyg+i ) and the length tl represent a line segment. Here, and ti are the above-mentioned values. Alternatively, it is also possible to use the maximum and minimum values of the X and X coordinates of a small area considered to represent a line segment as the coordinates representing the end point of the line segment.

In the above manner, the line segments As, ta, tt +ta, ts as shown in FIG. 6 are obtained from the original image shown in FIG.
.

tlo + tl2 r 415 Itll is calculated.

By the way, in the long sleeve and short axis lengths of the inertia equivalent ellipse obtained as described above, if a small area where the short axis length exceeds a certain value contains two or more line segments, It is assumed that there is. An example of this case is shown in FIG. 7. In such a case, the division into small regions as shown in FIG. 4 is performed again by narrowing the range of the angular difference in the direction of the concentration gradient. for example,
The area may be divided again using the minor axis direction of the inertial equivalent ellipse in this area as a threshold value in the concentration gradient direction, and the same processing as described above may be performed thereafter. However, the processing subsequent to this re-division into small areas does not need to be performed on the entire area of the target image, and may be performed only on small areas where the short axis length exceeds a certain value.

The procedure for line segment extraction according to this embodiment has been described above, but the above processing can be executed very efficiently by a general-purpose computer program or a dedicated image processing device. In particular, the processing from determining the density gradient of the original image to determining line segments as shown in FIG. 6 can be performed in one scan of the original image. The detailed processing procedure will be explained below with reference to FIG. First, the 0th, 1st, and 2nd moments for each small area, the maximum and minimum values of the X coordinate, the maximum and minimum values of the X coordinate,
A memory is prepared to store the maximum value, minimum value, etc. of the concentration gradient in the gradient direction. Next, each pixel of the original image is scanned from the bottom row to the bottom row, and in each row from the left end to the right end, the magnitude and direction of the density gradient are determined for each pixel, and if the magnitude exceeds a certain threshold, determines which subregion the pixel belongs to according to the range of the angular difference between the direction of the gradient of each pixel in the subregion to which the pixel directly above or under the pixel belongs and the direction of the gradient of the pixel. . In order to calculate the concentration gradient and divide the part with a large gradient into small regions according to the direction of the gradient, we need to calculate the pixel in one section immediately above the row to which the pixel belongs and the row in which the pixel belongs. For each pixel that has already been scanned, the number of the small area to which each pixel belongs may be stored. In addition, in each of the above-mentioned processes, it is only necessary to scan the entire original image once, and then all the values of the o-th, 1st-, and 2nd-order moments can be obtained by calculation, and furthermore, the above-mentioned calculations can be performed. The position of the line segment can be found.

As described above, in the present invention, p line segments are extracted by statistically processing not only information obtained from a thin region on the line but also information obtained from a wide region on the differential image. The reliability of extracted line segments is increased without being influenced by local noise.

Furthermore, since the equation of the line segment can be directly determined by a method such as determining the moment of a certain area of the image, it can be realized in a short time with a single process without requiring complicated processing.

Furthermore, unlike methods that perform processing while tracking a certain area, the equation of the line segment can be found by scanning the entire image a few times, reducing the amount of calculations and Hardware can also be easily realized.

[Brief explanation of the drawing]

Figures 1 to 8 relate to the present invention; Figure 1 shows the original image, Figure 2 shows the direction and magnitude of the grade gradient of the original image, and Figure 3 shows the density gradient. Figure 4 is a diagram showing a region where the size of Part 6 enlarged diagram showing the inertial equivalent ellipse
The figure shows extracted line segments, FIG. 7 shows a small area including two line segments, and FIG. 8 shows a screen during scanning. A...A region with a large concentration gradient, A1 to AI? ...Small area. Figure 1 Figure 2 Cause 3 Figure 4 Figure 4

Claims (6)

[Claims] (1) Divide the area in which the density gradient of the grayscale image exceeds a certain threshold into a plurality of connected small areas according to the range and connectivity of the density gradient direction, and for each small area, A line segment extraction method characterized by determining whether each small region is close to a line segment by determining second-order and second-order moments, and calculating the position of this line segment. (2) The line segment extraction method according to claim 1, wherein the 0th, 1st, and 2nd moments are weighted according to the magnitude of the concentration gradient. (3) The judgment as to whether the density gradient exceeds a certain threshold value and the selection of the range of the direction of the density gradient are made based on the sign of the differential value in two directions at each pixel point on the grayscale image and this differential value. 3. The line segment extraction method according to claim 1 or 2, wherein the line segment extraction method is performed using a magnitude relationship with a certain predetermined threshold value. (4) Detecting the magnitude and direction of the concentration gradient, determining whether the magnitude exceeds a threshold, dividing into small regions according to the range and connectivity of the concentration gradient direction, and each small region The line segment extraction according to any one of claims 1 to 3, wherein the calculation of the 0th, 1st, and 2nd moments in is performed in one scan of a grayscale image. method. (5) When determining the maximum and minimum values of the 0th, 1st, and 2nd moments and the XY coordinate values set on the grayscale image for each of the small regions, and determining that each small region is close to a line segment. , the line segment extraction method according to any one of claims 1 to 4, characterized in that the position of the line segment is determined using the maximum value and the minimum value. (6) Divide the region according to the range of the direction of the concentration gradient by determining whether or not it is close to a specific direction, and only the connected small regions whose concentration gradient is close to the specific direction are each close to a line segment. The line segment extraction method according to any one of claims 1 to 5, wherein only line segments in a direction perpendicular to the specific direction are extracted by determining whether or not .