JPH0812695B2 - Area extraction device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION [Outline] In a region extraction device, when a region occupied by an object in an image represented by an image information signal is extracted, edge detection means provides candidate points for the contour of the object. By extracting and assigning a code to the background area by applying the same code as the candidate point to the background area by means of background filling, and contracting the peripheral pixels of the false contour in the area to which the code is applied. Even if there is a shade change state, the area occupied by the object can be stably extracted.

[Industrial applications]

The present invention relates to a region extraction device, and more particularly, to a region extraction device adapted to extract a region occupied by an object in an image information signal whose density changes depending on light and shade.

[Conventional technology]

2. Description of the Related Art Conventionally, for example, a technique of extracting a region occupied by an observed object in an image signal obtained by a television camera or the like has been widely used.

For example, in an automatic device or the like that sorts a conveyed object on a belt conveyor according to its type, the above-described technique is essential. There, desired processing is performed based on the image signal from the television camera.

However, since the latest target objects are diverse,
It is necessary to distinguish and recognize a wide variety of objects. For example, even if there is a pattern on the surface of an object such as the cover of a book, a simple device configuration is desired that enables the book to be extracted as an object in an image signal.

As a region extraction method of such a device configuration, there is one as shown in FIG. In the figure, an image information signal 411 based on image pickup by a television camera (not shown)
Is converted into image information by the input unit 413 into a converted signal 415. Here, for example, a (512 × 512) pixel digital image is used. One pixel has 8 bits.

The threshold value signal 423 from the threshold value setting unit 421 is converted into a binarization unit.
It has been supplied to 431. In this binarization unit 431, the converted signal
For each pixel represented by 415, it is determined whether the pixel value exceeds a threshold value in the threshold signal 423. If it is higher, the pixel value is set to "1", and if it is lower, the pixel value is set to "0", and the binary signal 433 is generated.

In such a configuration, it is assumed that an object (not shown) placed on a background having a uniform color and brightness is imaged by a television camera and input as an image information signal 411. If the object also has a uniform color and brightness and is different from the background, the threshold signal 423
The object can be distinguished by the binarization conversion image processing in the binarization unit 431 with reference to the threshold of. That is, it is generally widely used because it is possible to determine which density portion belongs to the target area based on the difference in image density between the target and the background.

[Problems to be solved by the invention]

However, in the above-described conventional method, since it is based on the difference in image density between the background and the target, accurate area determination cannot be performed when the color and brightness of the surface of the target are not uniform.

For example, if there is a pattern on the surface of the object and the color and brightness have changed, the image density at some pixels may be lower than the background density. However, the concentration of the whole object is
If the density is higher than the background density, the pixel in the target object whose density is low will be judged as the background pixel by the binarization unit 431, and the target area cannot be extracted accurately. was there.

The present invention was created in view of such a point, and even if there is a pattern or the like on the surface of the object that changes the image density, the area of the image based on the object can be accurately extracted. It is an object of the present invention to provide a region extraction device configured as described above.

[Means for solving problems]

FIG. 1 is a block diagram showing the principle of the area extracting device of the present invention.

In the figure, when the image density gradient at each edge detection target pixel of a multi-valued digital input image signal in which each pixel is represented by multi-values exceeds a reference value, the edge detection means 120 sets the pixel to a first value, When the image density gradient in each edge detection target pixel is equal to or less than the reference value, a second value is assigned to the pixel and the multi-valued digital input image signal is output as a binary image.

The background filling unit 130 can be fixedly selected as a background pixel included in the multi-valued digital input image signal among the pixels having the second value of the binary image output from the edge detection unit 120. A binary image in which the pixel value of the pixel having the second value connected to the pixel is replaced with the first value is output.

The background area contraction unit 150 sets the pixel value of the pixel having the first value adjacent to the pixel having the second value in the binary image output from the background filling unit 130 to the second pixel value.
The binary image replaced with the value of is output.

The 2 output from the background area contraction means 150
The present invention is configured to extract an image region consisting of pixels having the second value in the value image.

[Work]

The candidate points forming the contour of the object in the image information output 111 are extracted by the edge detecting means 120, and the candidate points of the contour of the object are given a code (first value).

Then, the background filling means 130 fills each pixel in the region connected to the background pixel with the first value.

Then, the pixel value of the pixel having the first value adjacent to the pixel having the second value in the binary image output from the background filling means 130 is replaced with the second value by the background area contracting means 150. Then, the false contour that may be extracted in the image area to be extracted is erased.

In the present invention, the background is filled based on the extracted contour candidate points, and the false contour contraction processing (false contour erasing processing) is performed. Therefore, the pixel area to be extracted is confused with the background. Even if there is a density change state in which there is a shaded portion, the image area to be extracted can be accurately extracted without generating a contour in the density portion.

For example, the pixel value “1” is applied to the background, the contour of the object, and the pseudo contour inside the object by the background filling unit 130.
Is assigned. Then, after the filling processing, the background area shrinking unit 150 shrinks the area having a pixel value of "1", so that the contour of the object and the pseudo contour inside the object are erased, and only the background remains, and the area other than the background. Is extracted as a region of the object. If the background filling unit 130 is not provided, all the background is recognized by the shrinking process. Therefore, the two processes of the background filling process and the background region shrinking process should be combined to perform a meaningful process. Will be possible.

〔Example〕

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

FIG. 2 shows the structure of a region extraction device according to an embodiment of the present invention.

I, Correspondence between Embodiment and FIG. 1 Here, the correspondence between the embodiment of the present invention and FIG. 1 will be described.

 The image information signal 111 corresponds to the image information signal 411.

The edge detecting means 120 includes an input unit 413, an X-direction differentiating unit 211, and a Y-direction differentiating unit 211.
It corresponds to the direction differentiating unit 213, the calculating unit 231, the threshold setting unit 421, and the binarizing unit 431.

 The converted signal 113 corresponds to the binarized signal 241.

The background filling means 130 includes a thinning unit 253 and a contour point connecting unit 2
55, corresponding to background fill 257 and image memory 251.

The background area contraction unit 150 corresponds to the background area contraction unit 259 and the image memory 251.

The filled signal 121 corresponds to a signal exchanged between the image memory 251 and the background area contraction unit 259.

The area extraction output signal 141 corresponds to the signal given from the background area contraction unit 259 to the image memory 251.

II, Configuration and Operation of Embodiments Embodiments of the present invention will be described below on the assumption that there is a correspondence relationship as described above.

In FIG. 2, the same reference numerals as those in FIG. 4 denote the same circuit parts, and details thereof will be omitted here.

The image information in the image information signal 411 from the television camera is converted into a 512 × 512 × 8-bit digital image by the input unit 413. The converted signal 415 is
It is commonly supplied to the X-direction differentiator 211 and the Y-direction differentiator 213.

The X-direction differentiator 211 and the Y-direction differentiator 213 are spatial filters having shapes as shown in FIGS. 3A and 3B, and detect X-direction and Y-direction differential values of the input image. The X-direction differential signal 221 obtained by differentiating the converted signal 415 by the X-direction differentiator 211 and the Y-direction differential signal 223 obtained by differentiating the converted signal 415 by the Y-direction differentiator 213 are both calculation parts.
231 is supplied.

The arithmetic unit 231 takes the square root of the sum of squares of the differential value in the X direction by the differential signal 221 in the X direction and the differential value in the Y direction by the differential signal 223 in the Y direction, and supplies the calculated signal 233 to the binarizing unit 431. To do.

The threshold value setting unit 421 supplies a threshold value signal 423 representing a preset threshold value to the binarizing unit 431. This threshold value serves as a reference value for the differential value.

The binarizing unit 431 binarizes the operation signal 233 representing a digital magnitude by the threshold value of the threshold value signal 423. That is, if the value of the operation signal 233 is larger than the threshold value, “1” is output, and if it is smaller than the threshold value, “0” is output. The binarized information of each pixel in the binarized signal 241 thus obtained is stored in the image memory 251 at an address corresponding to each pixel.

Since the differential value takes a large value in the contour of the object, among the pixels stored in the image memory 251 by the binarized signal 241, the pixel of the “contour candidate point” based on the object is “1”.
Is assigned. Also, if there is a pattern inside the object, the pixel "false contour" in which the change in shading changes rapidly
Since the differential value also has a large value in “,” it is stored in the image memory 251 in a form in which “1” is assigned, and “0” is assigned to pixels other than “contour candidate point” and “false contour”. It was done.

In this way, the thinning unit 253 reads out the pixel information of “0” and “1” stored in the image memory 251 for all pixels, thins the area defined by the pixels of “1”, and then The result is stored again in the image memory 251.

The contour point connecting unit 255 sets “1” for the pixel adjacent to the “1” pixel among the information stored in the image memory 251.
And Such a changing process is repeated by the thinning unit 253 a predetermined number of times based on the contour line of the thinning result. Further, the changing process for setting the pixel adjacent to the pixel which is "0" to "0" is repeated the same number of times. By this, the discontinuity in the contour line is connected. The result is stored again in the image memory 251.

Based on the connected contour lines, the background filling part 257
Replaces the pixel "0" in the image memory 251 connected to the pixel at the upper left corner of the image with "1". The pixel in the upper left corner corresponds to "a pixel that can be fixedly selected as a background pixel" in the claims. Therefore, "fixed" in the claims means extraction of an image area. At
This means that it can always be used as a background pixel. And, "to be connected" in the "pixel connected to the pixel in the upper left corner" means that the pixel of "0" starts from the pixel in the upper left corner and continues to the connected pixel next to each other. Say.

After the replacement as described above, the replaced image is stored again in the image memory 251. This completes the background filling. By filling the background, the background, the contour of the object, and the pseudo contour inside the object can be extracted.

Further, the background area contraction unit 259 replaces the pixel adjacent to the pixel of the information “0” stored in the image memory 251 with “0”. By this process, not only the “contour candidate points” but also the pixels corresponding to the “false edges” inside the object are replaced with “0”, and only the pixels that are truly background have “1”. The contour of the object and all the points “including false edges” inside will have “0.” Then, it is stored again in the image memory 251.

In this way, the image memory 251 can obtain the information state representing the area of the object divided by the contour line.

III, Modification of Invention In the above-described embodiment of the present invention, the processing of the image information by the image pickup by the television camera has been described, but the present invention is not limited to this, and any image signal may be used. Good.

In addition, although the correspondence relationship between FIG. 1 and the embodiment of the present invention has been described in “I. Correspondence between Embodiment and FIG. 1”, the present invention is not limited to this, and there are various modifications. Of course.

〔The invention's effect〕

As described above, according to the present invention, it is possible to accurately extract a target contour area whose image density is changed in the contour from the image signal to be processed with a simple device configuration. Is extremely useful.

[Brief description of drawings]

FIG. 1 is a block diagram of the principle of the area extraction device of the present invention, FIG. 2 is a block diagram of the area extraction device of one embodiment of the present invention, and FIGS. 3 (A) and 3 (B) are FIG. 4 is an explanatory view of a filter of a differentiating part used in the embodiment of the present invention, and FIG. 4 is a configuration diagram of a conventional example. In the figure, 111 is an image information signal, 113 is a conversion signal, 120 is an edge detecting means, 121 is a filled signal, 130 is a background filling means, 141 is an area extraction output signal, 150 is a background area contracting means, and 211 and 213 are differentiating parts. , 221, 223 is a differential signal, 231 is a calculation unit, 233 is a calculation signal, 241 is a binarized signal, 251 is an image memory, 253 is a thinning unit, 255 is a contour point connection unit, 257 is a background filling unit, and 259 is a background. Region contraction unit, 411 is an image information signal, 413 is an input unit, 415 is a converted signal, 421 is a threshold setting unit, 423 is a threshold signal, 431 is a binarization unit, and 433 is a binarization signal. .

Claims (1)

[Claims] 1. When the image density gradient in each edge detection target pixel of a multivalued digital input image signal in which each pixel is represented by multivalues exceeds a reference value, a first value is assigned to the pixel and each edge detection target is detected. When the image density gradient in a pixel is equal to or less than a reference value, an edge detecting unit that assigns a second value to the pixel and outputs the multi-valued digital input image signal as a binary image, and an edge detecting unit that outputs the binary image Of the pixels having the second value of the binary image, the pixel value of the pixel having the second value fixedly connected as the background pixel included in the multi-valued digital input image signal And a background value filling means for outputting a binary image in which is replaced with the first value, and a first value adjacent to a pixel having the second value in the binary image output from the background filling means. Existence A background area contraction means for outputting a binary image in which the pixel value of a pixel is replaced with the second value, and the background area contraction means has the second value in the binary image output from the background area contraction means. An area extraction device characterized by being configured to extract an image area made up of pixels.