JPH0722326B2 - How to make a cutout mask - Google Patents

Description

Detailed Description of the Invention [Industrial applications]

The present invention relates to a cutout mask making method, and more particularly to a cutout mask making method suitable for use in making a printing plate-making film by cutting out a background portion from an input image.

[Prior art]

When you cut out and use only the necessary part from the photo original, for example, you can use it by trimming the square photo original into a heart shape, or you can extract only the object in the photo original and use it for other photo originals. When you use it by combining with,
Generally, cutout processing is executed. That is, trace the outline of the picture required by the trace machine on the layout paper to specify the cutout, cover the photo manuscript with tracing paper, draw the outline of the necessary part, and smash unnecessary photos with diagonal lines. Specify the cutout,
In accordance with such cutout designation, a cutout mask or the like is created and synthesis is performed. However, the printed matter is not limited to the one having the single color area as described above, but the color gradually changes, and there are negative and positive films for color photographs. When a film version is created by extracting only certain objects such as people or furniture, the areas to be cut out overlap with areas of different lightness, saturation, and hue. The film version cannot be created automatically. Therefore, in the case of a color film original, the film image is projected, a person creates a mask corresponding to the cutout area, and the mask is overlapped with the film original, so that the film plate of the desired color original ( Y , M,
Since it is left to manual work such as creating C and Bk plates, the film plate creation efficiency is significantly reduced due to the complexity of the image of the film original, and the number of printing process days is significantly increased. There was a problem. In order to solve such a problem, the background portion (background area) to be cut out in the image is designated, the color image data for the designated background portion is read, and the average density value and the plate color for each plate color are read. The covariance value between the plate colors is calculated, the color image data is read out for each pixel and for each plate color, and the discrimination function is used by referring to the density average value for each plate color and the covariance value between plate colors. It is conceivable to use the so-called maximum likelihood method of calculating the density distance data and extracting the background portion data while referring to the density distance data and the judgment distance data. When this maximum likelihood method is used, when the background is uniform, a cropped image can be created by designating a portion having the same pixel density, which is called a density mask. If the image is reflected or the background part has a gradation (may be related to lightness, saturation, and hue, respectively), you must specify the appropriate background parts at multiple places.
There is a problem that the operator is overloaded and an unskilled operator may not be able to specify an appropriate background area. In order to solve such a problem, there is a technique (Japanese Patent Laid-Open No. 1-298477) that attempts to achieve principal component analysis and automatic clipping by the maximum likelihood method based on the density distribution of a part of the background portion.
That is, the user determines the discriminant function by performing a principal component analysis for obtaining the axis of the density change based on the density distribution of the training area designated in advance in the background portion, and then
The user specifies a predetermined area including the contour of the target area, substitutes each data in the area into the previously obtained discrimination function, and binarizes it by the maximum likelihood method.

[Problems to be Solved by the Invention]

However, the above technique has a problem that the background portion apart from the first principal component axis of the training area may be erroneously discriminated toward the real portion side. For example, when a dark shadow is lying long and thin near the contour portion, it may be difficult to specify the training area so as to include this shadow while not covering the substantial portion. In such a case, the training area that does not include the shadow can be specified, and it can be determined to some extent by performing principal component analysis. However, the shadow color is the first in the training area distribution.
If it deviates significantly from the principal component axis, it may not be predictable. Further, in the case of an image in which the background portion includes two or more colors, the density distribution in the training area greatly expands, and thus the color on the substance portion side may be erroneously discriminated to the background portion side. The present invention has been made to solve the above-mentioned conventional problems, and an object of the present invention is to provide a cutout mask creating method capable of automatically and efficiently cutting out a cutout target image with a small number of discrimination errors. .

[Means for Solving the Problems]

According to the present invention, in a cutout mask creating method for arithmetically processing image data and electronically cutting it out, a training area extending over a background portion and a substantial portion in an image is set to 1
Specify one or more, statistically analyze the image data in the training area, the possibility that any color belongs to the background part,
Alternatively, an evaluation function indicating the possibility of belonging to the substantial part is created, a region that straddles the background part and the substantial part desired to be cut out of the image is specified, and all pixels in the specified region are specified by the evaluation function. The above object is achieved by substituting pixel data, determining whether the pixel belongs to the background portion or the substantial portion, and creating mask data from the pixels determined to belong to the background portion. Further, in the present invention, when creating the evaluation function, all pixel data in the training area are divided into two or more clusters, each divided cluster is displayed in a representative color, and the display color is changed to a background portion or a substantial portion. It is possible to select a cluster to be regarded and create an evaluation function for each cluster, which shows the possibility that an arbitrary color belongs to each cluster.

[Action]

In the present invention, one or more training areas extending over the background portion and the substantial portion of the image are designated, and the image data in the training area is statistically analyzed in, for example, a parameter space (color space), and an arbitrary color is determined. An evaluation function indicating the possibility of belonging to the background portion or the possibility of belonging to the actual portion is created, and by this evaluation function, an area (hereinafter referred to as a cutting area) that extends over the background portion and the actual portion where the image is desired to be cut out is created. For all pixels, it is determined whether the pixel belongs to the background portion or the substantial portion. Therefore, since the training area is specified over the background portion and the substantial portion, the evaluation function can be accurately determined, and the image to be cut out can be automatically and efficiently cut out with a small number of discrimination errors. To determine this evaluation function, for example, first, all pixel data in the training area is divided into two or more clusters.
This makes it possible to obtain clusters according to the distribution of pixel data. Then, each divided cluster is displayed in a representative color. As a result, processing corresponding to multi-valued pixel data can be performed. Next, a cluster to be regarded as a background part or a substantial part is selected from the display colors, and an evaluation function indicating the possibility that an arbitrary color belongs to the cluster is created for each cluster. Substituting each pixel data into this evaluation function and calculating, the closest cluster is selected. If the cluster is the background portion, it is determined that the pixel is the background portion, and if the cluster is the actual portion, the pixel is determined to belong to the actual portion and binarization is performed. Therefore, it is possible to accurately cut out an image in which the background portion and the substantial portion are composed of a plurality of colors. Especially suitable for cutting out objects with shadows. Further, since the training area and the area extending over the cutout line are separate areas, the training area can be made smaller. As a result, the clustering calculation amount can be suppressed, the processing time can be reduced, and efficient processing can be performed. Also, by dividing into clusters, it is possible to handle image data originally given in multidimensional form in multidimensional form. Therefore, the loss of information is less than that in the method of projecting onto a certain axis on the color space and treating it as a one-dimensional distribution and dividing into clusters. This enables highly accurate cutting.

【Example】

Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. This embodiment is a cutout mask forming apparatus having a structure as shown in FIG. FIG. 2 shows a conceptual diagram of the cutout mask forming apparatus viewed from the function. As shown in FIG. 1 and FIG. 2, the cutout mask forming apparatus includes three primary colors ( Y , M, C) from a color transparent original.
The image input means 10A for inputting the image data of the above, and the scanner 10 which is the mask output means 10B for outputting the processed image data on the film, and the input image or the output image is stored as necessary. A magnetic disk 12 for performing the processing according to the present invention to the input image to divide into a background portion or a substantial portion to create mask data in the background portion, and to output to the magnetic disk 12 or the scanner 10, a central processing A device (CPU) and memory unit 14, and an ON / OFF input means 16A for inputting ON / OFF information for clustering to the CPU and memory unit 14, and a training area or cutting A keyboard or digitizer 16 which is a means 16B for inputting area position information.
And a display unit 18 for displaying the image data being processed, the mask data after processing, or the representative color of each cluster on the screen.
And are provided. As shown in FIG. 2, the CPU and the memory unit 14 are for storing the image data storage memory 20 for storing the input image data and the position information of the training area input by the keyboard or digitizer 16. Training area storage memory 22 and the keyboard or digitizer
Of the image data in the cutting area storage memory 24 and the image data storage memory 20 for storing the positional information of the cutting area input in 16, the training area is divided into clusters and divided. In the clustering calculation means 26 for calculating a cluster table showing the number of clusters and information of each cluster, a cluster table storage memory 28 for storing the calculated cluster table, and the image data storage memory 20. Of the image data of, for the data in the cutting area,
It is provided with a binarization calculation means 30 for creating mask data by performing binarization calculation based on the cluster table storage memory, and a mask data storage memory 32 for storing the created mask data. Either the keyboard or the digitizer 16 may also serve as the on / off input means 16A and the position input means 16B, and each of them may be either the on / off input means 16A or the position input means 16B. Good. Since this embodiment is configured as described above, it has the following operation. The cutout mask creating apparatus of the embodiment executes the cutout mask creating process according to the procedure shown in FIG. When this procedure starts, first, in step 101,
Image data read by the scanner 10 or previously recorded on the magnetic disk 12 is input (loaded) to the CPU and the memory unit 14, and the image is displayed on the display unit 18. Next, the process proceeds to step 102, and it is selected whether the operation by the keyboard or the digitizer 16 is an operation of inputting the position designation of the training area or the cutting area, or an operation of selecting the cluster into either the background area or the substantial area. . However, at the beginning of this procedure, since the training area is first designated, only step 103 can be proceeded to. After the training area is designated and the processing of steps 104 and 105 is completed, at least 1
If the process proceeds to step 106 and the clusters are not selected, the process proceeds to step 107 and the position of the cutting area cannot be designated. In this procedure, cluster selection and training area designation can be changed in the course of processing. When the process proceeds to step 103, the position designation of the training area is input from the position input means 16A. In this training area, for example, as shown by reference numeral TA in FIG. 4, one or a plurality of locations extending over the substantial area and the background area in the image are designated. Next, in step 104, the input training area or areas are regarded as one area,
This is divided into clusters on the parameter space (color space of Y , M, and C axes in the embodiment). The details of this cluster division are performed by the procedure shown in the flowchart of FIG. 5, and as a result, the cluster table is output. The procedure of FIG. 5 will be described in detail later. The output cluster table stores the number of clusters and information on each cluster. The information of each cluster includes an average (vector) of data belonging to the cluster, a covariance matrix, the number of data (the number of pixels), and a flag indicating whether the cluster belongs to the background region or the substance region (for example, flag = “ 1 ”and does not belong to the flag =“ 0 ”). However, the value of this flag is not determined in this step 104, but is determined by the operator in the subsequent step 106. In the process of clustering, the clustering calculation means 26
A label image is temporarily used internally. This label image is a storage area for storing the number assigned to each cluster in association with the pixel data in order to indicate which cluster the pixel data in the training area belongs to. When it is determined that the pixel data belongs to a certain cluster, the number of the cluster is set as the label image of the pixel data. Here, regarding the pixel data in the training area,
The detailed procedure of the clustering calculation process will be described with reference to FIG. FIG. 5 (A) is a main routine for cluster division, and FIG. 5 (B) is a routine for dividing a cluster into two in the main routine. When dividing into clusters, as shown in FIG. 5 (A), first, in step 201, all the pixel data belonging to the training area are regarded as one initial cluster, and the average and covariance matrix of the data are calculated. To fill the cluster table. That is, the cluster table is initialized. At this time, the number of initial clusters is one. Next, in step 202, the label image is initialized by filling it with the initial cluster number (for example, 1). Next, the routine proceeds to step 203, where the routine for dividing the original cluster into two (cluster two division routine) shown in FIG. 5B is called. Since this cluster two-division routine is a type of routine that recursively calls itself in the routine, when the routine returns from the two-division routine to the main routine, the initial cluster has two clusters. It is divided into several clusters. When you call this routine, first step 301
Then, the initial value of the cluster center is determined for two new clusters (new clusters) formed by division. Each center of these two clusters is on the first principal component axis of the initial cluster (old cluster) as shown in FIG. 6, and the average value of these two centers is the average of the old cluster. Take an appropriate value to match the value (center). Note that the first principal component axis is the average of the old clusters in the color space,
That is, it is a straight line that passes through the center and has the eigenvector corresponding to the maximum eigenvalue of the covariance matrix as the direction vector, and is calculated based on the cluster table of the old cluster. Next, in step 302, each pixel data belonging to the old cluster is classified into two new clusters, and the label image is updated so as to be divided into 1 to 1 and 2, for example. In this classification, for each pixel data, an arithmetic expression for calculating a certain distance from each cluster is used, and the pixel data is classified (with a label) into the cluster having the smallest distance. The following (i) to (iv) can be considered as a method of calculating the distance d between the data and the cluster. (I) The Euclidean distance obtained by the following equation (1) on the parameter space is defined as the distance d. However, N: dimension of parameter space, k: cluster number, x = ▲ (xi) ^N _{i = 1} ▼; data, Ck = ▲ (Cki) ^N _{i = 1} ▼; at the cluster center (average) of the kth cluster is there. Using this Euclidean distance is equivalent to dividing the vertical bisector of the two new cluster centers as a boundary plane. (Ii) The Mahalanobis distance is set to the distance d as in the following equation (2). However, Co ▲ v ^-1 _k ▼ (i, j); the (i, j) component of the inverse matrix of the covariance matrix of the kth cluster. Note that the covariance matrix Cov is not determined in the first loop, so it is used from the second loop. (Iii) The sum of the distances in the respective axial directions is given by the following equation (3):
Let the distance be d. (Iv) Let the maximum value of the distance in each axial direction be the distance d as in the following expression (4). d (x, k) = max | xi-Cki | (4) Then, the process proceeds to step 303, where the position of the cluster center of the new cluster is updated. This update is performed by setting the average value of the pixel data classified into each new cluster as the new center of the cluster. Next, in step 304, it is determined whether or not to repeat the clustering process. In this case, the amount of movement of the cluster center updated in the previous step 303 with respect to the cluster center before update (temporary center) is calculated, and if it is a certain value or more, the process is repeated. After making a determination, the process returns to step 302 to repeat clustering. On the other hand, if it is less than the fixed value, it is determined that the repetition of clustering is finished, and the process proceeds to step 305. In step 305, the cluster table is updated. This update is performed by increasing the number of clusters by 1, replacing the information of the old cluster with the information of one of the new clusters, and writing the information of the other new cluster in the newly added cluster information field. Next, in step 306, the clustering end condition is determined for one of the new clusters. This determination is made, for example, when S and T are constants and det (Cov) <S is satisfied and the spread in the cluster is small, or (the number of data in the cluster) <T is satisfied and the number of data is small. Perform as a condition. If the result of the determination is that the termination condition is not satisfied, the operation proceeds to step 307, it is determined that the new cluster is still large, and there is room for the cluster, and the cluster two-division routine is recursively called to further divide the cluster. I do. On the other hand, if the end condition is satisfied, the process proceeds to step 308, and the clustering end condition of the other new cluster is determined in the same manner as step 306. If this end condition is not satisfied, the process proceeds to step 309 and step 307.
Similarly, the recursive call of the cluster division routine is performed to further divide the cluster. On the other hand, if the condition for ending clustering is satisfied, the process returns to the main routine for clustering shown in FIG. The cluster division is completed as described above. When the clustering is completed, the process proceeds to step 105 in FIG. In this step 105, each divided cluster is displayed on the display unit 18 in a representative color. A cluster center (average) color is used as a representative color. After completion of step 105, the process returns to step 102 and proceeds to step 106. In this step 106, the display
By comparing the representative color of each cluster displayed in 18 with the original image and designating whether the representative color is the background portion or the actual portion with the on / off input means 16A, each cluster is The flag in the cluster table is determined by designating whether it belongs to the background portion or the substance portion. Next, in step 107, the cutting area is input. Next, in step 108, the cutting area is binarized using the cluster table to create master data. That is, for each pixel in the cutting area, the distance between the color (pixel data) of the pixel and the center of each cluster is calculated from, for example, equations (1) to (4) (corresponding to an evaluation function), and the distance is calculated. The cluster having the smallest value is selected, and it is determined whether the cluster belongs to the background part or the real part.
If the cluster belongs to the background portion, the master data for the pixel is turned on, and if the cluster belongs to the substance portion, the mask data for the pixel is turned off. However, it is not necessary to calculate the distance d by the same calculation formula as that used in step 302, and the distance d may be different or the same, or the pixel mask on / You can calculate off. Next, in step 109, the created mask data is displayed, and in step 110, the mask data is stored in the storage memory 32. By the way, as a procedure for cutting out one image, first, the training area is input, then the on / off (on / off) of each cluster is input, and the cutting area is set along the contour line. It is possible to create masks by inputting them one after another. 1 around the image to be cut out
It is conceivable that various colors will appear before the lap. For example, in some places the color that belonged to the background area
At other locations, it can happen that it belongs to the real area. In order to deal with such a case, on / off of each cluster can be changed every time the cutting area is input once. When inputting cutting areas one after another, the mask on / off of each pixel in the overlapping portion is determined by, for example, the first
As a general rule, it is possible to give priority to the one calculated later, as shown in the table. In this case, if the desired contour line cannot be obtained, it can be corrected by changing the on / off of the cluster and then retaking the cutting area. The decision rule at the overlapping portion can be changed to an OR rule as shown in Table 2 and an AND rule as shown in Table 3 as required. In Tables 1 to 3, x is a mask value that has already been written, y is a newly calculated mask value, and z is a newly written mask value. Further, if the desired mask cannot be obtained by changing the on / off of the cluster, the training area can be retaken. For example, the quality can be improved by relocating the training area near the cutting area. Even in the above-described manner, a portion that cannot be cut out, for example, an extreme highlight portion or a shadow portion where the contour line originally does not appear can be manually retouched and corrected. As a correction method, a position to be corrected is specified from a position input means such as a keyboard digitizer and a mouse, and an area of a certain size centered on that position is masked on, and the specified area is masked off. As a method of erasing it, specify two point positions and mask between the two points on
The method of connecting with the line segment of 1 can be considered, and a mask can be created using either of them. By the end of the above processing, it is possible to make a pattern surrounding the pattern to be cut out with a mask. The unprocessed portion on the outside where the mask is created can be filled and filled with the mask. Further, smoothing processing can be applied to the jagged contour line, if necessary. The mask data is output by transferring the mask data to a plate collecting device and collecting the pattern together with other patterns, transferring the mask data to a scanner and exposing it on a film to form a mask plate, and contour line data as a vector. A method of obtaining a mask plate by converting the data into contours, transferring the contour line data to a cutting plotter and cutting the peel coat film, and the like can be considered. In the above embodiment, the C, M, and Y spaces are exemplified as the parameter space (color space) of the image data.
The space to be considered when carrying out the present invention is not limited to this, and in addition to this space, the color space of red (R), green (G), blue (B), and others, the differential value of each color plate It can be carried out by a combination or a calculation or conversion such as coordinates of luminance Y, chromaticity I, Q, hue H, lightness V, and saturation C.

【The invention's effect】

As described above, according to the present invention, it is possible to obtain an excellent effect that an image to be cut out can be automatically and efficiently cut out with a small number of discrimination errors. If the training area is divided into clusters with a variable number of clusters, first, multi-valued (quantization) of representative colors is performed according to the number of colors forming an image, and then binary It can be divided into a background part and a real part. As a result, it is possible to cut out an image in which the background portion and the substantial portion are composed of a plurality of colors. In particular, it is possible to accurately cut out a shadowed object. Further, since the training area and the cutting area are separate areas, the training area can be made small. As a result, the amount of calculation for clustering is suppressed, so that the processing time becomes short and the processing becomes efficient. In addition, in clustering, data originally given in multidimensional form can be handled in multidimensional form. Therefore,
The loss of information is less than the method of handling as a one-dimensional distribution projected on a certain axis and dividing into clusters. This enables highly accurate cutting.

[Brief description of drawings]

FIG. 1 is a front view including a partial perspective view showing a configuration of a clipping device according to an embodiment of the present invention, FIG. 2 is a block diagram showing a conceptual configuration of the device, and FIG. FIG. 4 is a flow chart showing a mask data creation procedure of the embodiment, FIG. 4 is a plan view showing an example of instruction of a training area in an image, FIG. 5 is a flow chart showing a detailed processing procedure of clustering, and FIG. FIG. 5 is a diagram showing an example of how to obtain a cluster center. 10 ... Scanner, 10A ... Image input means, 10B ... Image output means, 12 ... Magnetic disk, 14 ... Central processing unit (CPU) and memory section, 16 ... Keyboard or digitizer, 16A ... On / OFF input means, 16B ... position input means, 18 ... monitor (display means), 20 ... image data storage memory, 22 ... training area storage memory, 24 ... cutting area storage memory, 26 ... cluster division calculation Means, 28 ... Cluster table storage memory, 30 ... Binary calculation means, 32 ... Mask data storage memory.

Continuation of the front page (51) Int.Cl. ⁶ Identification number Reference number within the agency FI Technical display location G06T 7/00 (56) References JP-A-3-190469 (JP, A) JP-A-2-1076 (JP , A) JP-A-2-202679 (JP, A)

Claims (2)

[Claims] 1. A method for creating a cutout mask for arithmetically processing image data and electronically performing cutout processing, wherein one or more training areas extending over a background portion and a substantial portion in an image are designated, The image data in the training area is statistically analyzed to create an evaluation function that indicates the possibility that an arbitrary color belongs to the background portion or the actual portion, and the background portion and the entity desired to be cut out of the image are created. By specifying an area that spans a part and substituting the pixel data into the evaluation function for all pixels in the specified area, it is determined whether the pixel belongs to the background part or the substantial part, and the background part A method for creating a cutout mask, characterized in that master data is created from pixels determined to belong to. 2. The method according to claim 1, wherein when the evaluation function is created, all pixel data in the training area is divided into two or more clusters, each divided cluster is displayed in a representative color, and the display color is changed to the background portion. Alternatively, a clipping mask creating method is characterized in that a cluster to be regarded as a substantial part is selected, and an evaluation function indicating the possibility that an arbitrary color belongs to each cluster is created for each cluster.