JPS60169977A - Segmenting method of picture area - Google Patents

Abstract

PURPOSE:To extract a specific area in a picture rapidly without any specific skill by allowing a computer to detect the border of a noticed area highly accurately in accordance with human's simple instruction. CONSTITUTION:An objective picture 14 is AD converted by a picture input device 15 such as a drum scanner and stored in a storage device 18 through a central processor 16. The inputted picture data are displayed on a display device 19. While observing the display, the operator operates a coordinate input device 20 such as a track ball to designate plural representative points on an outline of the noticed area in the objective picture. The picture data 83 are stored in a picture memory 23 and displayed on a CRT monitor 26 through a DA converter 25. In this case, a cursor is displayed so as to be superposed to the picture. Receiving a command, a cursor generating device 24 sends a coordinate value indicating the cursor position to the central processor 16. Thus, picture coordinates of the plural representative points on the outline are successively applied to the central processor 16 by the operator's operation.

Description

[Detailed Description of the Invention] [Field of Application of the Invention] The present invention relates to computer processing of images such as photographs, and in particular to extraction of a specific area in an image quickly and accurately through simple interaction processing between a computer and a human. This invention relates to a suitable image area extraction method.

[Background of the invention]

Conventionally, in the process of cutting out only a specific area, such as a person, from an image such as a photograph and removing the remaining area, such as the background, a desired portion is cut out from photographic paper using a knife or the like. Alternatively, this was done by filling in the background using a brush and paint. Therefore, skill is required. A large amount of manpower is required to process the
The drawback was that it required time. Therefore, several attempts have been made to digitize images and perform area extraction processing using computers. The first person A connects an image display device to the computer, displays the target image there,
A person uses a coordinate input device such as a trackball or joystick to move a cursor superimposed on the image on the display along the boundary between the area of interest in the image and the background, thereby determining the coordinates of the boundary line. In this method, the computer uses this coordinate data to cut out the region of interest and erase the background. This method has the disadvantage that it requires time and skill to accurately indicate and input the outline of the region of interest. The second method is to have a computer automatically extract the region of interest, which examines the intensity value of each pixel in the target image using the difference in shading level or color difference between the region of interest and the background. This is to determine whether or not the area is inside the area.

However, this method has the disadvantage that the boundaries of the region cannot be detected with high precision unless the gray level or color is sufficiently uniform within the target region. In the third method, in order to detect sudden changes in pixel intensity at the boundary between the region of interest and the background, the pixel intensity is spatially differentiated for the entire image, and pixels whose absolute value of differential coefficient is greater than a certain threshold are detected at the boundary. This is a method of detecting it as a candidate. However, in this method, in addition to the boundary line of interest, discontinuous lines of gray level within the region of interest and within the background are also detected, so it is additionally necessary to extract only the desired boundary line from this processing result.

As mentioned above, in conventional methods, the task was to easily and quickly extract a specific area in one image.

[Purpose of the invention]

An object of the present invention is to provide a method that allows a computer to detect the boundaries of a region of interest with high precision according to simple instructions from a human, thereby enabling rapid extraction of a specific region in an image without any skill.

[Summary of the Invention] In order to achieve the above object, the present invention is characterized in that a rough contour line is first specified manually, and an accurate contour line is detected from the rate of change in image intensity in the vicinity of this contour line.

[Embodiments of the invention]

Here, the principle of area boundary detection in an image used in the present invention will be explained. The original image shown in FIG. 1 is the target image. Image 1 is a two-dimensional array of pixels, each having an intensity value. Consider a case where the region of the person 2 is extracted as the region of interest from among these and the other background region 3 is removed. FIG. 2 shows an image that should be obtained as a result of the desired processing.A region 2 of the person in the image 4 is the same as the original image, but the remaining background region 5 has the intensity of each pixel replaced by a predetermined intensity. That is, it is assumed that the background 3 in the original image is filled with uniform shading intensity. In order to realize this specific area extraction and background removal, the present invention extracts a specific area from the original image 1 as shown in FIG.

First, a contour line 6, which is the boundary between the target region of interest and the background, is extracted. Once this contour line is obtained, it is used to determine whether each pixel of the original image is inside the contour line 6, and if it is inside, the intensity of the corresponding pixel of the original image 1 is determined. A desired image 4 can be obtained by setting the given background intensity, if any, as the pixel intensity of the output image.

Next, a method for extracting the contour line 6 according to the present invention will be explained. As shown in FIG. 4, a plurality of individuals designate representative points on the outline of the region of interest in the original image 1. This is 7-1.7
-2. ...7-n (n total). This instruction is the input for the computer-based automatic contour extraction process described below, and high instruction accuracy is not necessarily required for the computer-based automatic extraction process. First, a straight line 8 connects consecutive representative points indicated above. This is called a rough contour line. Next, consider an inspection line segment 10 orthogonal to the rough contour line 8 at point 0 9, which will be explained using six representative points 9 as an example, for which the following processing is performed for each pixel on the rough contour line 8. line segment 1
The number of pixels between the zero end points 11 and 12 takes a predetermined value. The intensity of each pixel on this line segment 10 is inspected as follows, and a correct contour point, that is, a point 13 on the boundary line between the region of interest and the background is detected. FIG. 5 shows the intensity changes of pixels on the line segment 10. In this example, the pixel intensity of the region of interest 51 is large (that is, bright), and the intensity of the background region 52 is small (that is, dark). It can be seen from the pixel intensity that the boundary point is point 13. As shown in FIG. 6, this point is taken as the point where the absolute value of the differential value of the intensity change in the direction of line segment 10 in FIG. 5 is maximum.

By performing the above processing for each pixel point 9 on the rough contour line 8 and connecting the obtained points 13, a desired contour line can be obtained.

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

FIG. 7 is an overall configuration diagram of an apparatus for performing image processing according to the present invention. The target image 14 is AD converted by an image input device 15 such as a drum scanner, and then sent to a storage device via a central processing unit 16! Stored at 18. The input image data is displayed on the display device [19]. While viewing this display, the human operator operates the coordinate input device 20, such as a trackball, to indicate a plurality of representative points on the outline of the region of interest in the target image. This instruction does not necessarily have to be accurate; this operation is performed by a display device having the configuration shown in FIG. This is realized by 19. That is, the cursor generating device 24 follows the PIieA 81 and the command 82 input by the coordinate input device 20.

A cursor is displayed on the designated coordinates of the CRT monitor 26 via the DA converter 25.

On the other hand, image data 8 sent from the central processing unit 16
3 is stored in the image memory 23 and displayed on the CRT monitor 26 by the DA converter 25.

At this time, the cursor is displayed overlapping the main image.

The operator looks at the display on the monitor 23, operates the coordinate input device 20, and turns on the command switch 27 attached to the coordinate input device 20 when the cursor is positioned at a desired position. A command 82 is then sent to the display device 19. Upon receiving the command, the cursor generator 24 sends coordinate values indicating the cursor position to the central processing unit 16. In this manner, the image coordinates of a plurality of representative points of the contour are sequentially provided to the central processing unit 16 through human operations.

Then use these coordinates to centrally process the bag! ! 16 performs the processing shown in FIG. 9 to automatically extract the outline of the region of interest in the image. First, in step 91.

A rough contour line is created by connecting consecutive representative points indicated above with straight lines. In step 92, it is determined whether the intensity test has been completed for all pixels on the rough contour line, and if it has not been completed, in step 93, a test line segment perpendicular to the line is set for each pixel on the rough contour line. . The length of this line segment is assumed to be Ω pixels (Ω is about 10). In step 94,
The differential value (rate of change) of the intensity of each pixel is calculated along this line segment. In step 95, a process of 0 or more is performed for each pixel on the coarse contour line, in which the maximum point of the absolute value of this differential value is set as the contour point on the inspection line segment, and by connecting the obtained contour points, a desired contour line is formed. get.

Next, using this contour line data, the central processing unit 16 determines whether each pixel of the target image is inside the region of interest. For internal pixels, the pixel intensity of the original image is directly used as the pixel intensity of the output image. For pixels determined to be external, the pixel intensity value of the background area set by the operator from the console 17 is set as the intensity of the pixel in the output image.

As a result, desired output image data from which the background has been removed can be obtained. This data is sent to a display device w19 for display, and is used as an output image 22 by printing a photo onto a sheet of paper using an image output device such as a film writer [21].

Although it is possible to extract a specific area in an image with the above operation, depending on the image, the contour line extraction may be affected by noise in the image, and the obtained contour line 101 may have high-frequency irregularities as shown in Figure 10. Conceivable.

In that case, you can take one of the following steps.

(1) Smoothing is performed before step 94 in FIG. 9 (differentiation processing of pixel intensities on inspection line segments). Now let the index of the pixel on the line segment be i (i=1°..., 11) and the intensity be x (i). At this time, the simple differential value is d(i)-x(i)-xci-1)-(1), but instead of this, we first take the smoothed value x(i), but w(k
) (k=J-m+1.-, m) is determined by weight, and d(i)=x(i) x(i-1) (3) is used as a differential value.

(2) The obtained contour line 101 is subjected to two-dimensional Fourier transform to remove its high frequency components, and then subjected to inverse Fourier transform.

(3) Perform least square fitting of the obtained contour line 101 to a piecewise smooth curve. For example, the coordinates (xt y) are expressed as, and the coefficient an t bn (n=o, 1g...
.

N) using the least squares method.

A smooth contour line can be obtained using the above method.

On the other hand, if it is desired to blur the outline, such as the hair of a person, the following procedure can be used: In other words, a low-pass filter is applied to each pixel point of the outline determined by the method of this embodiment. The intensity of the pixel (it j) after region extraction is x(l,
j), the pixel near the contour line (1'#J'
) for M AX(1'#J') = Σ Σ w(kJ)x(i
'-, j'-n)k=-Mn=-M...(5) However, Is, w (k e Ω) performs a masking process which is a weight.

As described above, according to this embodiment, it is possible to automatically extract a specific area in an image with a simple human operation.

〔Effect of the invention〕

According to the present invention, a computer can automatically and highly accurately cut out a specific area in an image using outlines roughly specified by a human, so background removal, etc., which could only be done by an experienced person in the past, can be done in photos. This has the effect of reducing the cost of image processing.

In addition, the image cutting function, together with the operation function 2 editing function such as image enlargement/reduction, has the effect of increasing the effectiveness of document image processing by a computer.

[Brief explanation of drawings]

Figure 1 is the image to be processed, Figure 2 is the output image with the background removed, Figure 3 is the contour of the person in Figure 1, Figure 4 is a diagram showing the automatic contour extraction method, and Figure 5 is the output image with the background removed. FIG. 4 is a pixel intensity distribution diagram on the line segment 10, FIG. 6 is a diagram showing the rate of change in pixel intensity on the same line segment 10, FIG. 7 is an overall configuration diagram of the image processing apparatus according to the present invention, and FIG. 9 is a block diagram of an image display device constituting the apparatus, FIG. 9 is a processing flowchart of contour line detection, and FIG. 10 is a diagram showing the result of contour line extraction. Bud 4 Ward l6 /6 Rod 9 figure

Claims (1)

[Claims] An image region cutting method characterized in that an accurate contour line of a region of interest in an image is detected from a rate of change in image intensity using a roughly specified contour line.