JPH03201173A - Cutting processing method and device - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

[Detailed description of the invention] [Industrial application field]

The present invention relates to a cropping method and apparatus, and more particularly to a cropping method and fat suitable for use in electronically cropping using a cropping mask in order to extract a necessary substantial area from color image data.

[Conventional technology]

When creating a printing plate master film from color image data, generally a cropping mask is created for the unnecessary image area (background area), the background area is cropped using the mask, and the necessary substantive area is extracted. are doing. There are various techniques for creating this cropping mask. For example, input image data is displayed on a display screen, and an operator, while looking at the image on the screen, draws the outline of the image to be cropped. There is a technology to precisely instruct and cut out. There is also a technique in which an operator, while viewing an image displayed on a display screen, specifies the color of a background area that is desired to be masked, and automatically creates a cutout line based on the instruction. Also, while looking at the image displayed on the display screen, the operator specifies a part of the background area that is desired to be masked, finds the density change axis of that area, and then creates a density distribution based on the density change axis taking into account the density change axis. There is a technology that automatically creates cutout lines.

[Problem to be solved by the invention]

However, in the conventional technique of faithfully cropping according to the contour line specified by the operator, the operator needs to specify all the contour lines of the image using a cursor, etc., which takes time and puts a burden on the operator. The problem is that it is also large. In addition, with technology that automatically creates cutting lines according to color instructions or from the density change axis, if an area with a color similar to the background area is inside the solid area, that area may be accidentally cropped. In addition, there is a problem that even if there is a shadow near the outline, it may not be possible to cut through it well. As a technology related to the present invention, the applicant has already proposed an automatic clipping system in Japanese Patent Application No. 129997/1983. This automatic clipping system is similar to the present invention in that the background part is designated as a training area and an evaluation function is created taking into consideration the first principal precept axis (=direction of gradation). Therefore, this system also deals with gradations and shadows to some extent. However, the specific method for creating the evaluation function was developed, and there were points that needed to be improved. That is, in the system, no matter how far the gradation of the training area is extended, it is considered to fall within the background area, so that even a color far removed from the actual color of the training area may be determined to be the background area. Therefore, there is a risk that an unexpected color in the real area will be erroneously determined to be in the background area, and there is room for improvement. The present invention has been made in view of the above-mentioned conventional problems, and it is possible to quickly and accurately create a contour line as desired without distinguishing a color far from the color of the training area as a background area. An object of the present invention is to provide a clipping processing method and apparatus capable of extracting a substantial region image.

[Means to solve the problem]

The present invention is a method for electronically cropping image data by arithmetic processing, in which a part of the background area in an image is designated as a training area, and the image data within the training area is statistically analyzed for principal components. , create an evaluation function that indicates the possibility that an arbitrary color belongs to the background region by setting a tolerance limit on the magnification that extends in the direction of the principal component, specify the region that spans the cutout line of the image, and The above problem is solved by assigning the pixel data to the evaluation function for every pixel and determining whether the pixel belongs to a background area or a real area. The present invention also provides a clipping processing device having an image display means, an image data input means, and a calculation means, including a means for specifying a part of a background area in an image as a training area, and a means for specifying a part of a background area in an image as a training area. an evaluation function creation means for statistically analyzing the principal components of image data and creating an evaluation function indicating the possibility that an arbitrary color belongs to a background region by setting a permissible limit on the magnification for extending in the direction of the principal components; A means for specifying an area spanning the cutout line of an image, and substituting the pixel data into the evaluation function for all pixel data in the specified area, to which the pixel belongs to the background area or the real area. means for determining whether the pixels belong to the background area, means for creating mask data based on the pixels determined to belong to the background area, and means for writing the created mask data into memory and reading it out as desired. By providing the mask data storage means, the above problem is also solved.

[Effect]

In the present invention, a part of the background area in an image is designated as a training area, and the WfJ image data in the training area is statistically analyzed for principal components to show the possibility that any color belongs to the background area. Create an evaluation function by setting a permissible limit on the magnification that extends in the direction of the principal component, specify an area that spans the cutout line of the image, and apply the evaluation function to the evaluation function for all pixels within the specified area. By substituting the data, it is determined whether the pixel belongs to the background area or the real area. Therefore, in order to create an evaluation function from the image data in the training area by setting a permissible limit on the magnification extending in the direction of the principal component, the background up to the contour line outside the training area is calculated based on the pixel information in the training area. This means that the color is reasonably well estimated, and it is possible to create a more accurate evaluation function with less room for erroneous discrimination than the evaluation function proposed in Japanese Patent Application No. 129997/1988, which was previously proposed by the applicant. Since the evaluation function created in this way separates the real area and the background area, it is possible to accurately distinguish colors or shadows that are close to the background area, and also to identify colors that are far from the training area as background areas. There is nothing to do. Therefore, it is possible to quickly and accurately create a cutout line and extract a desired substantial area from an image. For example, if pixel data X is assigned to the evaluation function f(y), and the real constant do of positive real value is set as the threshold, if the assigned evaluation function value f(y)>d o, the pixel belongs to the real area. , f(y)≦do, it is determined that the pixel belongs to the background area. It is possible to create binary mask data in which mask data is turned on for pixels that are determined to belong to the background area and mask data is turned off for pixels that are determined to belong to the real area, and then a cutting process can be performed. Here, the evaluation function f(y) includes the Mahalanobis distance w
It is possible to use ld. The Mahalanobis distance d means a certain distance from a point of interest in a multidimensional space to a distribution in the space, and is based on how far the point of interest is from the distribution, taking into account the distribution state such as the distribution shape. 0 color space (
For example, cyan (C) axis, magenta (M) axis, yellow (
(Y) axis) is also a multidimensional space, and a certain distribution is formed when the data of each pixel in the training area is plotted in the color space, and the Mahalanobis equation between the distribution on the color space X and the element (pixel data) y is also a multidimensional space. The distance Md is calculated using the following formula (1) where Y is the average (vector) of elements y in the training area, and COV is a covariance matrix. For example, the Mahalanobis distance z in a -dimensional distribution is
This approximately corresponds to the deviation value δ as shown in the following equation (2). δ=Z* 10+50 (2) The Malahanobis distance in a one-dimensional distribution is expressed as Z=(x-n)/σ from the standard deviation σ, element x, etc., and is also called the Z value. In three-dimensional space, a collection of points with a constant Mahalanobis distance from the distribution forms an ellipsoid.Therefore, the method of using the Mahalanobis distance as an evaluation function and making a judgment based on a certain threshold value is, in other words, a method of sampling as a training area. For example, the color space distribution of the image shown in FIG. Figure 1 (B) shows the state in which the distribution is surrounded by an ellipsoid in which the square of the Mahalanobis distance Nd is 40.The training shown in Figure 1 (A) The average i is calculated from the areas, and the distribution of the training areas is shown in FIG. 1(B). However, if the Mahalanobis distance d is used as is in the evaluation function f (X), the image can be divided into a background region and a real region, but the following problem arises in further increasing the cutting precision. In other words, the training area is an area obtained by sampling a part of the background area, but not the entire background area, so
It is difficult to say that accurate evaluation is possible, and since the training area is an area where only a part of the background area is specified so as not to cover the actual area, it is necessary to specify the part of the outline between the actual area and the background area. Difficult,
Furthermore, in a pattern with shadows, colors other than those in the training area may appear near the outline. In order to improve this problem and improve accuracy,
The inventor conducted various studies and found that an evaluation function for attacking arbitrary elements (pixel data) in color space X can be determined as shown in the following equation (3). f(x)' t = (R'P-' (X -2)) A'x (R-1
P-' (y-2) ) ...... (3) However, there is a relationship of covariance matrix Cov = PAP-', and A is a diagonal matrix in which the eigenvalues of the matrix Cow are arranged in descending order,
P is an orthonormal matrix in which the eigenvectors for each eigenvalue of the matrix Cov are arranged horizontally as column vectors. or,
Assume that the vector y,7 is represented by a column vector,
The upper right subscript t in () represents transposition (for example, A
, and yt is a column vector converted into a row vector). Note that if n is the dimension of color space x, then R is a diagonal matrix expressed as the following equation (4), where 0 is a real constant of 1 or more, and is a real matrix of n n n. If ro=1, R becomes an identity matrix and the formula is (
3) becomes the Mahalanobis distance @d. This equation (4) calculates the ellipsoid when the Mahalanobis distance 11d is constant, for example, as shown in FIG.
This is equivalent to stretching it twice. The meaning of this enlargement is to know the direction of the gradation within the training area, and to allow a color that is an extension of that gradation as the background area. That is,
This equation (3) is based on the idea that if a color other than the training area appears near the contour line, there is probably already a strong gradation toward that color within the training area. ,(3)
If you evaluate with the formula, based on the data in the training area,
It has the effect of being able to predict its color. Here, if we consider that no matter how far the gradation of the training area extends, it falls within the background area, even colors that are far removed from the actual training area color may be determined to be the background color. There is a risk that an unexpected color within the area may be mistakenly identified as a background area. Therefore, in the present invention, an evaluation function is created by setting a permissible limit on the magnification ro extending in the direction of the principal component. This means that the background color up to the contour line outside the training area is reasonably estimated based on the information in the training area, and the evaluation function can be used to reduce or eliminate discrimination errors. I can create it.

【Example】

Embodiments of the present invention will be described in detail below with reference to the drawings. The system configuration of the cutting processing device according to this embodiment is shown in the third section.
As shown in the figure. As shown in FIG. 3, this system includes an image input unit 10 for inputting image data, an image data storage memory 12 for storing the input image data, and a training area and a cutout on the image data. A position input unit 14 for instructing an area spanning a line (hereinafter referred to as a cutting area), and a training area storage memory 16 and a cutting area storage memory for storing the input positions of the training area and cutting area, respectively. 18, and the evaluation function f of the above formula <3> from the stored image data and training area.
(y) and an evaluation function storage memory 2 for storing the created evaluation function.
2, and the evaluation function f(
x) by substituting the data of the image into x) and turning mask data on for pixels determined to be included in the background area using a threshold value do, and turning off mask data for other pixels to create binary mask data. 24, a mask data storage memory 26 for storing the created mask data, and a mask data output section 28 for outputting the stored mask data for cutting out the image data. This embodiment creates mask data according to the procedure shown in FIG.
It is stored in the storage memory 12 or displayed on a display means (not shown) (step 101). Next, the operator selects whether the position to be input from the position input unit 14 is the training area or the cutting area (step 102). ■At the beginning of this procedure, the evaluation function f (X)
Because it is necessary to create the training area in advance, only the training area can be entered.Also, after selecting the training area input, the cutting areas can be selected sequentially, and the training area can be changed as the cutting process progresses. is now possible. Then enter the training area (step 10)
3) In this input, it is possible to input, as desired, one or more training areas that include only the background part of the image data. Next, an evaluation function f (y) is created from the above equation (3) (step 104). This evaluation function f (y) is created by calculating the average color data (element) number of pixels in the input training area. 7. This is done by finding the covariance matrix COv, etc., and substituting it into equation (3). At this time, in addition to (red (R), green (G), blue (B)), the color spaces include (C, M, Y), (C1Y, M, K
), (luminance Y, chromaticity I, Q), etc. can be targeted. Note that when a plurality of training areas are designated in the previous step 103, evaluation functions f(y) are created corresponding to the number of training areas. Next, enter the cutting area (step 1)
05). Next, mask data is created from the image data within the cutting area (step 106). In this case, each pixel data X of the cutting area is converted into an evaluation function f(
x). Thereafter, the threshold value do is compared with the value of the evaluation function assigned to the image data. If there are multiple evaluation functions, each is compared. If at least one of the evaluation function values is less than or equal to the threshold do, the pixel is determined to belong to the background area and mask data is turned on for the pixel, and none of the evaluation function values is If the threshold value do is exceeded, the pixel is determined to belong to the real area, and the mask data is turned off. Furthermore, mask data is turned on to fill the outside of the background area to create binary mask data. The completed mask data is stored in the storage memory 26. Next, the mask data is displayed on a display means (not shown).
Step 107) If the displayed mask is inappropriate, the process returns to the light step 102 again to continue the process, and if it is appropriate, it is stored as mask data (Step 108). Appropriate mask data can be created in the manner described above, and the background region can be accurately cut out from image data using this mask data. In this case, even if a color other than the color in the training area appears near the contour line, an accurate cutout line can be created because equation (3) extends the Mahalanobis distance. In addition, since the tolerance limit is set at the magnification "0" in equation (3), a color that is far from the color in the training area is not determined to be the background color, and highly accurate mask data can be created. The area spans the background part and the real part, and mask data can be created regardless of whether the instruction includes or does not include the training area.In the above embodiment, automatic cutting processing is performed. Although the embodiment has been described, when implementing the present invention, the manual cutting method may be combined with this method to carry out the cutting process.

【Effect of the invention】

As described above, according to the present invention, it is possible to obtain the excellent effect that the cutting process can be performed accurately and quickly without identifying a color that is far from the color of the training area as a background area.

[Brief explanation of drawings]

FIG. 1 is a plan view and a line diagram showing an example of a training area and its pixel data distribution surrounded by Mahalanobis distances for explaining the present invention in detail; FIG. Fig. 3 is a block diagram showing the overall configuration of the cutout mask creation system according to the embodiment of the present invention, and Fig. 4 explains the operation of the embodiment. 3 is a flowchart showing a procedure for creating a cutout mask for. 10... Image input section, 12... Image data storage section, 14... Position input section, 16... Training area storage memory, 18...
Cutting area storage memory, 20...Evaluation function creation unit, 22...Evaluation function storage memory, 24...Binarization calculation unit, 26...Mask data storage memory, 28...Mask output means.

Claims (2)

[Claims] (1) In a method of computationally processing and electronically cropping image data, a part of the background area in the image is designated as a training area, and the image data within the training area is statistically analyzed for principal components; Create an evaluation function that indicates the possibility that an arbitrary color belongs to the background area by setting a tolerance limit on the magnification for extending in the direction of the principal component, specify the area that spans the image cutout line, and calculate all within the specified area. 1. A clipping processing method, comprising: substituting the pixel data into the evaluation function to determine whether the pixel belongs to a background area or a real area. (2) In a cropping processing device having an image display means, an image data input means, and a calculation means, a means for specifying a part of the background area in the image as a training area, and a means for specifying the image data in the specified training area. An evaluation function creation means for statistically performing principal component analysis to create an evaluation function indicating the possibility that an arbitrary color belongs to a background area by setting a permissible limit on the magnification for extending in the direction of the principal components; means for specifying an area spanning the cutout line; and for all pixels within the specified area, assigning the pixel data to the evaluation function to determine whether the pixel belongs to a background area or a real area. means for creating mask data based on pixels determined to belong to the background area; and mask data storage means for writing the created mask data into a memory and reading it out as desired. A cutting processing device characterized by comprising: and.