JPH1079019A - Method for analyzing picture - Google Patents

Abstract

PROBLEM TO BE SOLVED: To easily detect a shading area, and to facilitate recognition of the shape of a curved surface by operating picture analysis by utilizing the change in a differentiating direction. SOLUTION: A non-maximum value is eliminated by performing one- dimensional differentiation to an original picture having object 1 and 2, and then an extended picture processing is operated according to a result obtained by comparing it with a threshold value so that a step-shaped edge can be extracted. Next, a shading area candidate is extracted, and a shading area is extracted. Finally, the step-shaped edge below a constant value in the shading area including the step-shaped edge is deleted as one part of the shading area, and basically separated as a shading area wave step-shaped edge. Thus, the step-shaped edges of the object 1 and 2 are extracted from the original picture by a threshold processing using a differentiating method, the shading area candidates of the object 1 and 2 are extracted according to the change of the differentiating direction of each point of the picture, and the extracted step- shaped edges are compared with the shading area candidates so that the shading areas of the objects 1 and 2 can be extracted.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image analysis method for detecting a shading area or a curved area of an object from an original image.

[0002]

2. Description of the Related Art When an image is structurally analyzed, it is necessary to extract a region having the same property of the image or to detect an edge where the property of the image changes rapidly. In other words, the image is divided into regions of uniform brightness to extract the area occupied by the object, or the edge of the object, which is a change point of a feature such as brightness, is detected and the contour of the object is detected. Extraction is necessary for recognition of image contents.

Conventionally, such image properties include brightness / color uniformity of image and texture (textur).
e) It focuses only on the property. When a curved surface object is targeted, the presence or absence of shading is also important information.

[0004]

In the conventional image analysis, however, there is a problem that it is difficult to extract a shading area and to recognize a curved surface shape because no extraction method has been established. Was.

The present invention has been made in view of the above problems, and has as its object to provide an image analysis method which can easily extract a shading area and can easily recognize a curved surface shape. I have.

[0006]

An image analysis method according to the present invention comprises the steps of: extracting a step-like edge of an object from an original image by threshold processing using a differentiation method; And extracting a shading region of the object by comparing the extracted step-shaped edge with the shading region candidate.

In the step of extracting the shading area candidates, the differential intensity and the differential direction of each point of the image are calculated by differential processing, and the image is binarized based on the differential intensity.
A hole filling process is performed for a region having a small and small differential intensity, and a removing process is performed for a region having a large and small differential intensity.

In the step of extracting the shading area, the shading candidate area and the step-like edge are overlapped, and the shading candidate area is reduced by deleting pixels having a large differential intensity except pixels overlapping the step-like edge. After the processing, a region having a large differential intensity with an area equal to or smaller than a certain value is removed, and the remaining region having a large differential intensity is subjected to expansion processing.

In addition, a curved surface shape of the extracted shading area is detected.

[0010]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An image analyzing method according to the present invention comprises the steps of extracting a step-like edge of an object from an original image by threshold processing using a differentiation method, and shading the object from a change in the differential direction of each point of the image. The method includes a step of extracting a region candidate and a step of comparing the extracted step-shaped edge with a shading region candidate to extract a shading region of the object.

Here, the shading area can be defined as follows. That is, it does not include the step-like edge obtained by the conventional differentiation method, but has a differential value on a constant value, and a connected region having a small change in the differential direction, having a size equal to or larger than a certain value. Can be defined.

FIG. 1 is a process chart showing an image analysis method for extracting the shading area. Hereinafter, the extraction process of the shading area will be described step by step with reference to FIGS. 2 to 9.

(1) First, extraction of a step-like edge (S
1) The original image having the objects 1 and 2 shown in FIG. 2 is subjected to first-order differentiation to eliminate non-maximal values, and then subjected to extension image processing based on a result of comparison with a threshold value, thereby obtaining an image shown in FIG. Such a step-like edge is extracted.

(2) Extraction of shading area candidates (S2
-S5) (2a) Next, the differential intensity and the differential direction of each point of the image as shown in FIG. 4 are calculated by the same first-order differentiation processing as when extracting the step-like edge.

(2b) The image is binarized as shown in FIG. 5 with the point where the differential intensity is equal to or greater than a certain value set to “1” and the other points set to “0”.

(2c) The small "0" area is regarded as lacking in the shading area, and its filling processing is performed.

(2d) Also, as shown in FIG. 6, the small "1" region is removed (to "0") as noise.

(3) Extraction of shading area (S6-S
10) (3a) Next, the extracted shading area candidate and the step edge are superimposed.

(3b) For the shading area candidate,
As shown in FIG. 7, the reduction processing is performed a plurality of times (here, four times). At this time, the pixel of “1” is deleted, but the pixel of “1” overlapping the step edge is not deleted.

(3c) The region of "1" having an area smaller than a certain value is removed as shown in FIG.

(3d) The remaining area of "1" is expanded and restored. That is, the pixel which is set to “1” in the expansion processing is made to match the pixel before the processing of the above (2).

(4) Finally, of the shading area including the step-like edge, the step-like edge having a certain value or less is deleted as a part of the shading area. However, basically, as shown in FIG. Separate with step edges.

By the above-mentioned processes (1) to (4), the shading area mainly appears on the curved surface, and is hardly detected on the plane. Therefore, by detecting the shading area, the possibility of the existence of the curved surface becomes clear, and the shape of the curved surface can be known by stereo vision between the shading areas.

As described above, the differential value is equal to or more than a certain value,
Among the connected regions of which the change in the differential direction is small and the direction is the same, a region having a certain size or more can be easily extracted as a shading region, and its curved surface shape can be easily recognized. .

Originally, the differentiation method is used to detect a step-like edge where the brightness of an image changes suddenly as a large differential value portion. However, as in the present embodiment, a change in the differentiation direction is examined. It is also possible to extract a shading area where the brightness of the image changes slowly.

Here, the above-mentioned differential direction is a brightness gradient direction represented by the following equation when SX = h−b and SY = f−d with respect to the neighboring pixels shown in FIG.

Tan ^-1 (SY / SX) The differential intensity is expressed by the following equation.

[0028]

(Equation 1)

As shown in FIG. 11, the filling process is to bring the chipped portion into the same state as the surroundings.

The reason why the reduction process is performed four times is 1 depending on the object.
This may be performed five times or more, but the number of pixels that can be reduced at one time is only one nearby pixel, so that noise cannot be reduced to the state shown in FIG. 7 unless repeated a plurality of times. It should be noted that the dotted lines in the figure indicate step-like edge portions, and the black portions indicate shading regions.

The expansion process is performed to restore the reduced image as shown in FIG. 12 while removing noise.

The present invention can be applied to not only an on-vehicle camera, but also equipment that recognizes the outside world and performs various operations based on the data, and controls a robot operating on a production line or the like in a factory. It can also be applied to equipment.

[0033]

As described above, according to the present invention, since the image analysis is performed using the change in the differential direction, the shading area can be easily detected, and the recognition of the curved surface shape can be easily performed. There is an effect that there is.

[Brief description of the drawings]

FIG. 1 is a process diagram showing an image analysis method according to the present invention.

FIG. 2 is a diagram showing an original image.

FIG. 3 is a diagram showing a step-like edge;

FIG. 4 is a diagram showing a differentiation direction.

FIG. 5 is a diagram showing a state after threshold processing based on differential intensity.

FIG. 6 is a diagram showing a state after a hole filling and a noise removal processing;

FIG. 7 is a diagram showing a state after reduction processing;

FIG. 8 is a diagram showing a state after a small area deletion process.

FIG. 9 shows a shading area.

FIG. 10 is an explanatory diagram showing neighboring pixels.

FIG. 11 is an explanatory diagram illustrating an example of a filling process.

FIG. 12 is an explanatory diagram showing an example of noise removal.

[Explanation of symbols]

 1 object 2 object

 ────────────────────────────────────────────────── ─── Continued on the front page (72) Inventor Fumiaki Tomita 1-1-4 Umezono, Tsukuba, Ibaraki Pref. Electronic Technology Research Institute (72) Inventor Eiji Takagi 2-21-, Kasuga, Tsukuba, Ibaraki 13 Tsukuba Plaza 503 (72) Inventor Yutaka Ishiyama 1-3-1 Edanishi, Aoba-ku, Yokohama-shi, Kanagawa Prefecture Stanley Electric Co., Ltd.

Claims (4)

[Claims] A step of extracting a step-like edge of an object from an original image by threshold processing using a differentiation method; a step of extracting a shading region candidate of the object from a change in a differentiation direction of each point of the image; An image analysis method, comprising a step of comparing a step-shaped edge and a shading area candidate to extract a shading area of the object. 2. The step of extracting a shading area candidate includes calculating a differential intensity and a differential direction of each point of an image by a differential process, binarizing the image based on the differential intensity, and then calculating an area of a small and small differential intensity. 2. The image analysis method according to claim 1, further comprising performing a filling process and removing a region having a large and small differential intensity. 3. The step of extracting a shading area includes superimposing a shading candidate area and a step-like edge,
After the shading candidate area is reduced by deleting pixels having a large differential intensity except pixels overlapping with the step-shaped edge, a region having a large differential intensity with an area equal to or less than a certain value is removed, and the remaining differential is removed. 2. An expansion process for an area having a large strength.
Or the image analysis method according to 2. 4. The image analysis method according to claim 1, wherein a curved surface shape of the shading area is detected.