JPH05101184A - Automatic setting method for interesting area - Google Patents

Abstract

PURPOSE:To obtain the method which can set the interesting area of high quality with high reproducibility. CONSTITUTION:A template image wherein the background and object are completely separated is inputted 2, the contour of the object is extracted 3 by image processing, and an auxiliary contour which is moved by a constant distance from the contour to the inside of the object is found 4; and plural interesting area elements are arranged 5 along the auxiliary contour and the sum set of those interesting area elements is set as the interesting area of the template image. The respective interesting area elements arranged along the auxiliary contour which is offset to the contour by the specific quantity include the contour (border line of density variation), so the interesting area of the template image set as the sum set of the respective interesting area elements effectively includes the density variation, and so the interesting area has good quality and highly precise template matching becomes possible.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method of automatically setting a region of interest in image processing.

[0002]

2. Description of the Related Art In image processing, the degree of similarity (or dissimilarity) between a template and an image is determined by determining in which part of the image a pattern (hereinafter, referred to as a template) of an object registered in advance exists. A method called template matching for finding the position of the target object may be used.

In the template matching, as shown in FIG. 6, the template area S is superimposed on the target images (bearing images) a, b, c ... And the target images overlapping the template A and the template area S (in the illustrated example, A method of calculating the degree of similarity (or dissimilarity) between the target images b) and determining the value as the probability that the object at that position exists (or the probability that the object does not exist). It is performed for all points in the image, and the position where the similarity is maximum (or the dissimilarity is minimum) or the position where the similarity exceeds a certain threshold (or the dissimilarity is less than a certain threshold) is targeted. It is said to be the position of the object.

By the way, when the objects are flat such as bearings as in the example shown in FIG. 6, the objects are often stacked. In such a case, in order to perform image processing based on the feature amount, it is necessary to separate the object by some method. This is because it is necessary to correctly grasp the outer shape of the object on the image in order to extract the feature amount, otherwise the subsequent processing cannot be performed correctly.

However, in practice, it is not easy to cut out the objects, and at present, the objects are supplied one by one by a parts feeder or the like, or the objects supplied from a parts maker are manually rearranged step by step. The mechanical method of is adopted.

Therefore, a separation method using a region of interest has been proposed as a method of separating the background and the shape of an object without resorting to the mechanical method. In this method, a region to be image-processed (region of interest) and a region not to be subjected to image-processing (non-region of interest) are determined in advance with respect to an image input for image processing, and image processing is performed accordingly. Is a way to do.

[0007]

However, in the conventional separation method using a region of interest, the template image is unconditionally divided into mosaic blocks of 4 × 4 pixel units when setting the region of interest, and each block is divided into blocks. If the density is constant, the block is a non-interest region, and if the density is not constant, it is a region of interest. Therefore, it is not possible to effectively capture the density change points included in the template, and it is of good quality. There was a problem that the region of interest could not be obtained.

The present invention has been made in view of the above problems, and an object thereof is to provide an automatic region-of-interest setting method capable of setting a high-quality region of interest with good reproducibility.

[0009]

In order to achieve the above object, the present invention takes in a template image in which a background and an object are completely separated, extracts the contour line of the object by image processing, and Obtain an auxiliary contour line that has moved a certain distance from the line toward the inside of the object, arrange a plurality of ROI elements along the auxiliary contour line, and set the union of these ROI elements as the ROI of the template image. The feature is that it is set.

[0010]

According to the present invention, the contour line of each object extracted by the image processing constitutes the boundary line of density change, and is arranged along the auxiliary contour line offset by a predetermined amount with respect to this contour line. Since each region-of-interest element that has been created includes a contour line (border line of density change), the region of interest of the template image set as the union of each region-of-interest element effectively contains density changes, and Will be of high quality and highly accurate template matching will be possible.

[0011]

An embodiment of the present invention will be described below with reference to the accompanying drawings.

FIG. 1 is a flow chart showing the processing procedure of the method of the present invention, FIG. 2 is a diagram showing an input image for a template, FIG. 3 is a diagram showing contour lines of an extracted object, FIG. 4 and FIG.
FIG. 4 is an enlarged detail view of a portion X of FIG. 3, FIG. 4 is a view showing auxiliary contour lines determined along the contour line, and FIG. 5 is a view showing region-of-interest elements arranged along the auxiliary contour lines. is there.

According to the method of the present invention, a template image in which the background and the object are completely separated is taken in, the contour line of the object is extracted by image processing, and a constant distance from the contour line toward the inside of the object. Find the auxiliary contour that has moved only by
A plurality of ROI elements are arranged along the auxiliary contour line, and the union of these ROI elements is set as the ROI of the template image.

Now, the method of the present invention will be described in detail when the object is a bearing.

First, the offset amount ε of the auxiliary contour line from the contour line is preset (step 1 in FIG. 1), and then the template input image as shown in FIG. 2 is taken in (step 2 in FIG. 1). ). In this input image, as shown in FIG. 2, the background B and the target bearing C are completely separated in two colors, black and white, by the binarization process.

Next, the template input image is used to perform image processing for subtracting the template input image from the actual input image of the bearing C on a pixel-by-pixel basis. The ring-shaped contour lines C1, C2, C3 and C4 are extracted (step 3 in FIG. 1). Then, an auxiliary contour line which is moved from the extracted contour lines C1 to C4 toward the inside of the bearing C by the offset amount ε is obtained (step 4 in FIG. 1). For example, as for the contour lines C1 and C2, the auxiliary contour lines C1 ′ and C2 ′ are obtained by moving the contour lines C1 and C2 in the direction of the illustrated arrow by an offset amount ε as shown in FIG.

As described above, the auxiliary contour lines C1 ', C
When 2 '... Is obtained, the auxiliary contour lines C1', C2 '...
A plurality of region-of-interest elements e are arranged along the line (step 5 in FIG. 1). For example, as for the auxiliary contour lines C1 'and C2', a plurality of region-of-interest elements e are arranged along the auxiliary contour lines C1 'and C2' as shown in FIG. Then, the union Σe of these ROI elements e is set as the ROI of the template image.

By the way, when the region of interest is set as described above, Hadamard transformation (for details, see Japanese Patent Laid-Open No. 2-83786) is applied to all the region of interest elements e constituting the region of interest. In step 6) of FIG. 1, the Hadamard coefficients in each region of interest element e are calculated, these Hadamard coefficients are stored in the memory as a template (step 7 in FIG. 1), and thereafter, template matching is performed using this template. ..

Thus, in this embodiment, the contour lines C1 to C1 of the bearings C, which are the objects extracted by the image processing, are selected.
C4 constitutes a boundary line of density change, and the contour lines C1 to C
Auxiliary contour line C offset by a predetermined amount ε with respect to 4
Since each region-of-interest element e arranged along 1 ′, C2 ′, ... The area will effectively contain density changes,
The region of interest has high reproducibility and high quality, and highly accurate pattern matching is possible.

By resetting the offset amount ε when determining the auxiliary contour lines C1 ', C2' ..., It is possible to easily perform the trial setting of the regions of interest having different tendencies.

[0021]

As is apparent from the above description, according to the present invention, the template image in which the background and the object are completely separated is taken in, the contour line of the object is extracted by image processing, and the contour is extracted. Obtain an auxiliary contour line that has moved a certain distance from the line toward the inside of the object, arrange a plurality of ROI elements along the auxiliary contour line, and set the union of these ROI elements as the ROI of the template image. Since it is set, it is possible to obtain an effect that a high quality region of interest can be set with good reproducibility.

[Brief description of drawings]

FIG. 1 is a flowchart showing a processing procedure of a method of the present invention.

FIG. 2 is a diagram showing a template input image.

FIG. 3 is a diagram showing a contour line of an extracted object.

FIG. 4 is a diagram showing auxiliary contour lines determined along a contour line.

FIG. 5 is a diagram showing region-of-interest elements arranged along an auxiliary contour line.

FIG. 6 is an explanatory diagram of template matching.

[Explanation of symbols]

 B Background C Bearing (object) C1 to C4 Contour line C1 ', C2' Auxiliary contour line e Region of interest element

Claims (1)

[Claims] 1. A template image in which a background and an object are completely separated is taken in, an outline of the object is extracted by image processing, and the inside of the object is moved by a certain distance from the outline. An automatic region-of-interest setting method comprising: obtaining an auxiliary contour line, arranging a plurality of ROI elements along the auxiliary contour line, and setting a union of these ROI elements as a ROI of a template image.