JPH06233315A - Area separator for color iamge signal - Google Patents

Abstract

PURPOSE:To accurately separate an area for the change of a light source and the noises by enabling the setting of a subtle threshold value. CONSTITUTION:The color video signals are converted into the hue, saturation and lightness signals. The luminance signal is stored in a background storage part 2, and the hue and saturation signals undergo the color conversion through a color converting part 4 based on a color model of a subject. A lightness difference detecting part 5 detects the difference between the lightness signal acquired by conversion and the stored lightness signal. Thus a probability image generating part 6 generates a probability image from the acquired color signal and lightness difference signal. Then the probability signal is binarized by an area separating part 7, and only the area of the subject is separated and transmitted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a region separating apparatus for image signals, and more particularly to a target object in a color image when only the video of the target object in the video is cut out and the object is measured or recognized. The present invention relates to a region separating device that extracts only a region.

[0002]

2. Description of the Related Art In the field of image processing, there are cases in which only the image of a target object is cut out from the image signal, and the object is measured or recognized, or combined with another image signal. Conventionally, there are three methods described below as methods for extracting a target object from a color image. First,
As a first method, the color video signal is the three primary colors of light,
RGB consisting of three components, R (red), G (green), and B (blue)
The second method is to express the R, G and B components by H (hue), S (saturation),
Converted to an HSV representation consisting of three components of V (brightness),
A method of thresholding the value of HS and a third method are to convert it into an HTV expression composed of three components of H (hue), T (color intensity) and V (lightness), There is a way to threshold the value.

[0003]

However, in the first method described above, since the three primary colors are used as they are, when a change in brightness occurs due to a change in illumination or a shadow of a target object,
Since the value of the color component also changes greatly, there is a drawback that the target object is erroneously recognized. In addition, the second and third methods have a drawback that the area separation accuracy is lowered by individually thresholding HS or HT.

Therefore, the main object of the present invention is to perform noise processing and threshold processing on an image, to enable delicate threshold setting, and to accurately separate light source changes and noise. An object of the present invention is to provide an area separating device for such a color image signal.

[0005]

An area separation device according to the present invention receives a color image signal for displaying a plurality of subjects in a predetermined area, and outputs a color image signal for each of hue, saturation and lightness. First conversion means for converting, background storage means for storing a brightness signal obtained after conversion of a background image signal from which a plurality of subjects have been removed by the first conversion means, and a color model of the target object before or during startup. A color model creating means for creating a color image signal, a second converting means for performing color conversion of each of the hue and saturation signals obtained by converting the color image signal by the first converting means using the color model, and a color image signal The third conversion means for detecting the difference between the signal obtained after conversion by the first conversion means and the brightness signal stored by the background storage means, and the respective signals obtained by the second and third conversion means. A fourth converting means for creating Luo probability image constituted by a separation means for separating only the fourth region of the target object by binarizing the probability signal obtained by the conversion means.

[0006]

The target object separating apparatus according to the present invention creates a color model of the target object from the average and variance of the hue and saturation of the target object, and uses this color model to calculate the hue and color of the color image signal. The degree signal is converted into a color probability image.
On the other hand, the brightness signal of the image signal is converted into a brightness probability image from the difference from the brightness signal stored in the background storage means.
After that, the product between the color probability image and the brightness probability image is taken and converted into a probability image of the object. The conventional threshold processing is performed before the probability image of the object is created.
It was difficult to accurately separate the values because they were binarized. However, in the present invention, since a probability image is used and noise processing and threshold processing are performed on this image, it is possible to set a delicate threshold, and the light source Accurate separation is possible for changes and noise.

[0007]

1 is a schematic block diagram of an embodiment of the present invention, and FIG. 2 is a block diagram of the color model creating section shown in FIG.

First, referring to FIG. 1, the area dividing device
It includes a color model creating section 3 for creating a color model S7 of the target object and an arithmetic processing section 9 for receiving the color image signals S1 and S7 and performing area separation arithmetic processing. The arithmetic processing section 9 includes an image conversion section 1 for converting the color video signal S1 into signals of hue, saturation and lightness, and a switching section 8 for selecting the image signals S2 and S3 generated from the image conversion section 1. The image signal S2 is a signal for displaying only the background not including the target object, and the image signal S3 is a signal for displaying an image including the target object. The image signals S2 and S3 are given to the background storage unit 2 or the color conversion unit 4 via the switching unit 8. The background storage unit 2 receives the image signal S2 and stores only the brightness signal S2 ₂ among them. The color conversion unit 4 receives the color-only signal S3 ₁ of the image signal S3 provided via the switching unit 8, and obtains the color probability image signal S4 ₁ using the color model signal S7. The brightness difference detection unit 4 receives the brightness-only signal S3 _{2 of the} image signal S3 including the target object provided via the switching unit 8 and the brightness signal S2 ₂ from the background storage unit 2, and outputs the brightness probability image signal S4 ₂ . obtain. The probability image creating unit 4 receives the color probability image S4 ₁ and the brightness probability image signal S4 ₂ , and obtains a probability image signal S5 of the object. Then, the area separation unit 7 receives the probability image signal S5 and obtains the area signal S6 of the target object.

FIG. 2 is a schematic block diagram of an example of the color model creating section 3 shown in FIG. With reference to FIG. 2, first, the color model creating section 3 includes an image converting section 31 for converting the color video signal S31 into a color signal S32 ₁ and a lightness signal S32 ₂ . The image storage unit 32 stores the brightness signal S32 ₂ and simultaneously outputs the brightness signal S33 ₂ of the previous frame. The brightness difference detection unit 35 uses the brightness signal S3 of the current frame.
2 ₂ and the brightness signal S33 _{2 of the} previous frame are received, and a difference signal S34 between them is obtained. The color model conversion unit 33 extracts the color information of the target object region obtained from the difference signal S34 by using the color signal S32 ₁ to create a color model S7.

FIG. 3 shows an example of a color model, FIG. 4 shows an example of obtaining a probability distribution using the color model of FIG. 3, and FIG. 5 shows an example of a highly accurate color model. 6 is a diagram showing an example in which a probability distribution is obtained using the color model shown in FIG.

Next, referring to FIG. 1 to FIG.
The split processing will be specifically described. First, the image conversion unit 1
Receives the color video signal S1. Usually one image
The video signals that are generated are frames that are temporally continuous.
You will get an image, but the following explanation will simplify the explanation.
For one image, that is, one frame of image signal
The processing will be described. The image conversion unit 1 is given 1
After video signal S1 for the screen is A / D converted, H (hue), S
(Saturation) and V (brightness) signals are separated. First,
The brightness signal of only the background that does not include the target object is stored.
Therefore, the switch of the switching unit 8 is switched to B, and the brightness signal S2₂
Only the background is stored in the background storage unit 2. Next, the switch unit 8 switch
Switch to A to switch the color conversion unit 4 to the target object.
Hue and saturation signal S3 including body₁Give brightness difference detection
The brightness signal S3 including the target object to the unit 5 ₂To send. Color conversion
In the part 4, the color signal S3 is calculated using the color model S7.₁Strange
The color probability image S4₁Is output.

For example, assume that the color model 7 is that shown in FIG. The hue is one period from the origin to hx, the saturation takes a value from the origin to the maximum value sx, and the existence probability of the target object is 0 at the origin and 1 at the maximum value px. Further, it is assumed that the color signal is given as shown in FIG. If all the saturation values in the area 1 are s1 or more and only the hue values in the area 2 are h1 or more and h2 or less, the output color probability image is the area 2 and the other binary images. . For example, the saturation value on the line segment ab is shown in FIG.
4B, and when the hue value is as shown in FIG. 4C, the output value is as shown in FIG. 4D. This is equivalent to the conventional processing in which the hue and the saturation are independently binarized and the product of them is taken. On the other hand, in the case where the color model is as shown in FIG. 5,
The probability distribution on the output line segment ab is as shown in FIG. 6 (d), which shows that much information is stored as compared with FIG. 4 (d).

The brightness difference detection unit 5 takes the difference between the brightness signal S3 ₂ containing the target object and the brightness signal S2 ₂ of the background storage unit ₂ to create a brightness probability image signal S4 ₂ . In the probability image creating unit 6, the color probability signal S4 ₁ and the lightness probability image signal S
The product of 4 ₂ and the existence probability image of the target object are output as S5.

FIG. 7 is a diagram showing an example of the probability distribution of the probability image S5. In particular, FIG. 7 (a) shows the distribution obtained by using the above method, and FIG. 7 (b) shows the hue and saturation. , Brightness difference, thresholded each image, and ANDed 2
It is a value distribution. It can be seen that the distribution is very sensitive to noise in each image. On the other hand, in FIG. 7A, it can be seen that the noise between the images is canceled out.

In the area separation unit 7, the probability image as shown in FIG. 7B is binarized after applying a noise elimination filter such as a low frequency component pass filter. Since the binarization can be performed immediately before the area signal S6 of the target object is obtained,
It is easy to adaptively change the threshold for the value.

Next, referring to FIG.
The method will be specifically described. Color video signal S31
Image conversion unit 31 converts each signal of hue, saturation, and brightness.
Replace. Brightness signal S32₂Is recorded by the image storage unit 32.
At the same time, the brightness signal S33 of the previous frame is stored.₂Is output
Be done. The brightness difference detection unit 35 uses the brightness signal S3 of the current frame.
Two₂And the brightness signal S33 of the previous frame₂And receive them
To obtain the difference signal S34. In the color model conversion unit 33, the difference
With the value of the signal S34 as the weight W (i, j), the color saturation image S
32₁Hue and saturation values are sampled from. Service
The average value of the sampled hue and saturation
m_n, M_sAnd the variance value is σ_n, Σ _sTo
And, as an example, the probabilistic model P_{(h, s)}Is expressed as
You can

[0017]

[Equation 1]

[0018]

As described above, according to the present invention, a color model of the target object is created from the respective average values and variances of the hue and saturation of the target object, and the color model of the color image signal is created using this color model. The color and the saturation signal are converted into a color probability image, the brightness signal of the image signal is converted into a brightness probability image from the difference between the brightness signal stored by the background storage means, and the image is obtained from the color probability image and the brightness probability image. Since the product of the two is taken and converted into a probability image of the object, it is possible to set a delicate threshold value and accurately separate light source changes and noise.

[Brief description of drawings]

FIG. 1 is a schematic block diagram of a region separation device showing an embodiment of the present invention.

FIG. 2 is a block diagram of a color model creation unit shown in FIG.

FIG. 3 is a diagram showing an example of a color model.

FIG. 4 is a diagram showing an example of obtaining a probability distribution using the color model shown in FIG.

FIG. 5 is a diagram showing an example of a highly accurate color model.

FIG. 6 is a diagram showing an example of obtaining a probability distribution using the color model shown in FIG.

FIG. 7 is a diagram showing an example of a probability image.

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 image conversion unit 2 background storage unit 3 color model creation unit 4 color conversion unit 5 lightness difference detection unit 6 probability image creation unit 7 region separation unit 8 switching unit 9 arithmetic processing unit 31 image conversion unit 32 image storage unit 33 color model conversion Part 35 Brightness difference detection unit

Claims (1)

[Claims] 1. An area separation device for a color image signal, which separates the background portion from a color image signal obtained by capturing a plurality of subjects in a predetermined background and separates and extracts only the target subject. A first conversion unit that receives the color image signal and converts it into signals of hue, saturation, and lightness; and a background image signal from which the plurality of subjects have been removed, obtained after conversion by the first conversion unit. Background storing means for storing the brightness signal, a color model creating means for creating a color model of the subject before or at startup, and a hue and a saturation obtained after the color image signal is converted by the first converting means. Second conversion means for color-converting each of the signals using the color model, a lightness signal obtained after conversion of the color image signal by the first conversion means, and the background memory. Third conversion means for detecting a difference from the brightness signal stored by, fourth conversion means for creating a probability image from each signal obtained by the second and third conversion means, and the fourth conversion means. Area separating apparatus for color image signals, which comprises binarizing the probability signal obtained by the converting means to separate only the area of the target object.