JPH03211966A - Cut-off mask preparing method - Google Patents

Abstract

PURPOSE:To automatically and efficiently cut off the object picture with small discrimination error by designating more than one training area over a background part and a real body part in the picture and statistically analyzing picture data in the training areas. CONSTITUTION:A cut-off mask preparing device composed of a scanner 10, magnetic disk 12, central processing unit and memory part 14, keyboard or digitizer 16 and display part 18 designates more than one training areas over the background part and the real body part in the picture. The picture data in the training areas are statistically analyzed by a parameter space, for example, and an evaluation function is prepared to show possibility for an arbitrary color to belong to the background part or to the real body part. Concerning all picture elements in the areas over the background part and the real body part desired for the cut-off picture, it is judged whether the picture elements belong to the background part or the real body part. Thus, since the training areas are designated over the background part and the real body part, the evaluation function is accurately determined and the objective picture of cut-off is automatically and efficiently cut off with the small discrimination error.

Description

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

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

[Conventional technology]

When cutting out and using only the necessary parts from a photographic manuscript, for example, trimming a square photographic manuscript into a heart shape and using it, or extracting only the object from a photographic manuscript,
When a photo is to be combined with other photo originals and used, a cutting process is generally performed. In other words, use a tracing machine to trace the outline of the required image on a layout sheet, designate a cutout, cover the photographic manuscript with tracing vapor, draw the outline of the necessary part, and then cross out unnecessary photographs with diagonal lines. to specify the crop,
According to such clipping specifications, a clipping mask or the like is created and composition is executed. However, printed matter does not only consist of a single color area as mentioned above, but also includes gradual color changes, negative and positive films for color photographs, etc. When creating a film version by extracting only a specific object, such as a person or furniture, the area to be cut out overlaps with areas with different brightness, saturation, and hue, so the above method cannot produce a faithful image. Film version cannot be created automatically. Therefore, in the case of a color film original, a person projects the film image, creates a mask corresponding to the cutout area, and overlaps the mask with the film ia to create the desired film version of the color original ( Y, M, C
, Bk plate>, the complexity of the film original image significantly reduces the efficiency of film plate creation and significantly increases the number of days required for the printing process. There was a problem. In order to solve this problem, we specify the background part (background area) to be cut out in the image, read out the color image data for the specified background part, and calculate the density average value and plate color for each plate color. The color image data is read out for each pixel and each plate color, and the discrimination function is calculated while referring to the density average value for each plate color and the covariance value between the plate colors. It is conceivable to use the so-called maximum likelihood method, in which density distance data is calculated by , and background data is extracted while referring to the density distance data and determination distance data. When using this maximum likelihood method, when the foreground is -like, it is possible to create a cropped image by specifying areas with the same pixel density, which is called a density mask. If the background part has a gradation (it may be related to brightness, saturation, and hue), you must specify multiple appropriate background parts.
This method imposes a burden on the operator, and an unskilled operator may not be able to specify an appropriate background area. To solve this problem, there is a technique (Japanese Patent Laid-Open No. 1-298477) that attempts to achieve automatic cropping using principal component analysis and maximum likelihood method based on the density distribution of a portion of the background area. The discriminant function is determined by performing principal component analysis to determine the axis of density change based on the density distribution of a training area specified in advance in the background, and then the user specifies a predetermined area containing the outline of the target area. , each data within that area is substituted into the previously determined discriminant function and binarized using the maximum likelihood method.

[The problem that the invention tries to solve! ! ! ! ]however,
The above technique has the problem that the background part away from the first principal component axis of the training area may be erroneously identified as the real part. However, it is sometimes difficult to designate a training area that includes this shadow while not covering the real part. In times like this,
A certain degree of culm can be determined by specifying a training area that does not include the shadow and performing principal component analysis on it, but if the color of the shadow deviates significantly from the first principal component axis of the distribution of the training area, the prediction cannot be made. There are things I can't do. Furthermore, in the case of an image in which the background part includes two or more colors, the density distribution in the training area is greatly expanded, so that there is a possibility that the color on the real part side will be mistakenly determined to be on the background part 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 method for creating a cropping mask that can automatically crop an image to be cropped efficiently and with fewer discrimination errors. . [Means for Solving the Problems] The present invention provides a method for creating a cutout mask for calculating and electronically cutting out image data, in which a training area spanning a background part and a real part in an image is divided into one part.
Specify one or more colors and statistically analyze the image data in the training area to determine the probability that any color belongs to the background.
Alternatively, create an evaluation function that indicates the possibility of belonging to the real part, specify an area spanning the background part and the real part from which you want to cut out the image, and apply the evaluation function to all pixels within the specified area. The above problem is achieved by substituting pixel data, determining whether the pixel belongs to either the background part or the substance part, and creating mask data from the pixels determined to belong to the background part. . Further, in the present invention, when creating the evaluation function, all the pixel data in the training area is divided into two or more clusters, each divided cluster is displayed with a representative color, and the background part or the real part is distinguished from the displayed color. It is possible to select clusters to be considered and create an evaluation function for each cluster that indicates the possibility that an arbitrary color belongs to each cluster.

[Effect]

In the present invention, one or more training areas spanning the background and substance parts of an image are specified, and the image data in the training areas is statistically analyzed, for example, in a parameter space (color space), and an arbitrary color is Create an evaluation function that indicates the possibility of belonging to the background part or the possibility of belonging to the substance part,
Using this evaluation function, it is determined whether all pixels in an area spanning the background and substance parts of the image (hereinafter referred to as a cutting area) where it is desired to be cut out belong to the background part or the substance part. Therefore, by specifying the training area across the background part and the real part, the evaluation function can be determined with high accuracy, and the image to be cropped can be automatically cropped efficiently and with fewer discrimination errors. To determine this evaluation function, for example, first, all pixel data in the training area is divided into two or more clusters. Thereby, clusters can be obtained according to the distribution of pixel data. Next, each divided cluster is displayed using a representative color. This allows processing equivalent to multi-value conversion of pixel data to be performed. Next, a cluster to be considered as a background part or a real 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 it selects the closest cluster. If the cluster is in the background part, it is determined that the pixel is in the background part, and if the cluster is in the real part, it is determined that the pixel is in the background part. If there is, it is determined that the pixel belongs to the real part and binarization is performed. Therefore, it is possible to accurately cut out an image in which the background portion and the substance portion have a plurality of colors. Especially suitable for cutting out objects with shadows. Furthermore, since the training area and the area spanning the cutout line are separate areas, the training area can be made smaller. As a result, the amount of calculation for clustering can be suppressed, processing time can be reduced, and efficient processing can be performed. Furthermore, by clustering, image data originally given in multiple dimensions can be handled as it is in multiple dimensions. Therefore, compared to a method of projecting onto a certain axis on a color space, handling it as a one-dimensional distribution, and dividing it into clusters, there is less loss of information, which enables highly accurate cutting.

[Embodiment 1] Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. This embodiment is a cutout mask production apparatus having a configuration as shown in FIG. FIG. 2 shows a conceptual diagram of this cutout mask creation device viewed from the functional point of view. As shown in FIGS. 1 and 2, this cutout mask creation device uses three primary colors (Y, M, C) from a color transparent original.
A scanner 10 serves as an image input means 10A for inputting image data of , and also serves as a mask output means 10B for outputting processed image data onto a film, and a scanner 10 for storing input images or original images as necessary. A magnetic disk 12 for processing the input image according to the present invention to create mask data for the background or substance portion, and a central disk for outputting to the magnetic disk 12 or scanner 10. A processing unit (CPU) and memory section 14, and the CP
U and memory section 14 have on/off settings for clustering (
onloff) On/off input means 16A for inputting information and means 16B for inputting position information of the training area or cutting area.
A keyboard or digitizer 16, and a display section 18 for displaying image data being processed, mask data after processing, or a representative color of each cluster on a screen are provided. The CPU and memory section 14, as shown in FIG.
An image data storage memory 20 for storing input image data; a training area storage memory 22 for storing training area position information input using the keyboard or digitizer 16; Cutting area storage memory 2 for storing positional information of the cutting area
4, among the image data in the image data storage memory 20, those in the training area are divided into clusters, and the number of clusters is calculated for calculating a cluster table showing the number of divided clusters and information on each cluster. A calculation means 26, a cluster table storage memory 28 for storing the calculated and determined cluster table, and a cluster table storage memory for storing image data in the cutting area among the image data in the image data storage memory 20. It is provided with a binarization calculation means 30 for performing binarization calculation based on the data and creating mask data, and a mask data storage memory 32 for storing the created mask data. Either the keyboard or the digitizer 16 may serve as both 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 too. Since this embodiment is configured as described above, it has the following effects. The cutout mask creation device of the embodiment executes the cutout mask creation process according to the procedure shown in FIG. When this procedure starts, first, in step 101, image data read by the scanner 10 or recorded in advance on the magnetic disk 12 is transferred to the CPU and memory unit 14.
input (load) the image into the display means 18. Next, the process proceeds to step 102, in which the keyboard or digitizer 16 is used to select whether the operation is to input the position designation of the training area or the cutting area, or to sort the clusters into either the background area or the real area. . However, at the beginning of this procedure, the training area is specified first, so it is only possible to proceed to step 103, and the training area is specified and step 104.
After the process in step 105 is completed, unless the process proceeds to step 106 at least once to select clusters,
It is not possible to proceed to step 107 and specify the position of the cutting area; however, in this procedure, it is possible to change the cluster sorting and the training area specification in the process of proceeding with the process. If the process proceeds to step 103, the position input means 16A
Enter the training area location from here. As this training area, for example, as shown by reference numerals T and A in FIG. 4, one or more locations extending between the real region and the background region in the image are designated. Next, the process proceeds to step 104, in which the input training area or multiple locations is regarded as one region, and this is divided into clusters on the parameter space (in the embodiment, the color space of the Y, M, and C axes). . The details of this clustering are performed according to the procedure shown in the flowchart for the fifth country, and a cluster table is output as a result. The procedure shown in FIG. 5 will be explained in detail later. The number of clusters and information about each cluster are stored in the output cluster table. Information on each cluster includes the average (vector) of the data belonging to that cluster, covariance matrix, number of data (number of pixels), and a flag indicating whether the cluster belongs to the background area or the real area (for example, flag = " 1” indicates that the flag belongs, and flag=“O” indicates that the flag does not belong). However, the value of this flag is not determined in this step 104 and is determined in the subsequent step 106.
This is determined by the operator. In addition, in the process of clustering, the clusters are calculated by calculation means 2.
6. Temporarily use the label image internally. This label image is a storage area for storing numbers assigned to each cluster in correspondence with pixel data to indicate which cluster the pixel data in the training area belongs to. When it is determined that pixel data belongs to a certain cluster, the number of that cluster is used as the label image of that pixel data. Here, regarding the pixel data in the training area,
The detailed steps of the calculation process for clustering will be explained with reference to FIG. FIG. 5(A) is a main routine for clustering, and FIG. 5(B) is a routine for dividing the main routine into two clusters. In clustering, as shown in FIG. 5(A), first, in step 201, the entire pixel data belonging to the training area is regarded as one initial cluster, and the average and covariance matrix of that data are calculated. to fill in the cluster table. That is, the cluster table is initialized. In addition, at this time,
The number of initial clusters is 1. Next, the process proceeds to step 202, where the label image is initialized by being completely filled with the initial cluster number (for example, 1). Next, the process proceeds to step 203, as shown in FIG. 5(B).
Routine for dividing the original cluster into two (cluster 2
split routine). Note that this two-cluster division routine recursively calls itself (recurs-ive call).
Since this is a type of routine, when the main routine returns from this two-division routine, the original cluster is divided into two or more clusters. When this routine is called, first step 301 is executed.
Then, the initial value of the cluster center is determined for the two new rasters (new clusters) created by the division. For example, as shown in Figure 6, the centers of these two clusters are on the first principal component axis of the original cluster (old cluster), and the average value of these two centers is the average value of the old cluster. Take an appropriate value so that it matches the value (center). Note that this first
The principal component axis is a straight line in the color space that passes through the mean or center of the old cluster and whose direction vector is the eigenvector corresponding to the maximum eigenvalue of the covariance matrix, and is calculated based on the cluster table of the old cluster. It is. Next, the process proceeds to step 302, where each pixel data belonging to the old cluster is classified into two new clusters, and the label image is updated, for example, from 1 to 1 and 2. This classification uses an arithmetic expression for calculating a certain distance between each pixel data and each cluster, and classifies (labels) the pixel data into a cluster having the smallest distance. The method of calculating the distance @d between data and cluster is as follows (
i) ~ (iV > etc.) (i) Let the Euclidean distance obtained as in the following equation (1) on the parameter space be the distance d. d(x, k)'=Σ(X 1- Ck i ) 2; tl (1) where, N: dimension of parameter space, k: cluster number, (average). Using this Euclidean distance is equivalent to dividing the perpendicular trisection plane between the two new cluster centers as the boundary surface. <ii) The Mahalanobis distance is expressed as the following equation (2): Let the distance be d. (XjCkj)...(2) However, cOV-i (i, j) is the (i, j) component of the inverse matrix of the covariance matrix of the first cluster. In the second loop, the covariance matrix Cov is not determined, so it is used from the second loop. (iii) The sum of the distances in each axis direction is the distance d, as in the following equation (3). d( x, k)=ΣIX t Ck il...(3);
zl (iV) Let the maximum value of the distance in each axis direction be the distance d as shown in the following equation (4). d (X, k) = nax 1x 1-Ck
1l-(4) Next, 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 that cluster. Next, the process proceeds to step 304, where it is determined whether or not to repeat the clustering process. The amount of movement is calculated, and if it is greater than a certain value, it is determined that the process is to be repeated, and the process returns to step 302 to repeat clustering; - if it is less than a certain value, clustering is performed. It is determined that the repetition of the above is completed, 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 information of one of the new clusters, and writing the information of the other new cluster in the newly added cluster information column. Next, the process proceeds to step 306, where the termination condition for one new cluster is determined. This determination is performed, for example, with S and T as constants and det(Cov
)<S holds and the spread within the cluster is small, or (number of data in cluster)<T holds and the number of data is small. As a result of the determination, if the termination condition is not satisfied, the process proceeds to step 307, where it is determined that the new cluster is still large and there is room for the cluster, and the cluster division routine is recursively called to further divide the cluster into clusters. I do. On the other hand, if the termination condition is satisfied, the process proceeds to step 308, and the determination of the termination condition for the other new cluster is performed in the same manner as in step 306. If this termination condition is not satisfied, the process proceeds to step 309. Proceeding to step 307, the cluster division routine is recursively called to perform further cluster division. On the other hand, if the termination condition for the cluster is satisfied, the process returns to the main routine for cluster division as shown in FIG. 5(A). . Clustering is completed in the above manner. When the clustering is completed, the process proceeds to step 105 in FIG. 3. In this step 105, the center (average) color of the cluster is used as the 0 representative color for displaying each divided cluster in the representative color on the display section 18. After this step 105 is completed, step 102
The process returns to step 106. In this step 106, the representative color of each cluster displayed on the display unit 18 is compared with the original image, and the on/off input means 16A is used to specify whether the representative color is the color of the background part or the substance part. By doing this, it is specified whether each cluster belongs to the background part or the real part, and the flag in the cluster table is determined. The process then proceeds to step 107, where a cutting area is input.Then, the process proceeds to step 108, where the cutting area is binarized using the cluster table to create mask data. That is, for each pixel in the cutting area, calculate the distance between the color (pixel data) of the pixel and the center of each cluster using, for example, the above formulas (1) to (4) (corresponding to the evaluation function), and calculate the distance. The cluster with the minimum 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 part, the mask data for the pixel is turned on, and if the cluster belongs to the substance part, the mask data for the pixel is turned off. However, it is not necessary to calculate the distance Md using the same calculation formula as that used in step 302; it may be different, the same, or (1) to (4).
Onloff of a pixel mask can be calculated from an evaluation formula outside the armor. Next, the process proceeds to step 109 to display the created mask data, and in step 110 this mask data is stored in the storage memory 32. By the way, the procedure for cropping one image is to first input the training area, then input the on/off status of each cluster, and then,
It is conceivable to create a mask by inputting cutting areas one after another along the contour line. It is conceivable that various colors appear during one rotation around the image to be cut out. For example, it may happen that a color that belongs to the background area at one location belongs to the solid area at another location. To deal with this case, each time you enter the cutting area, the onlo of each cluster is
ff can be changed. When inputting cutting areas one after another, the mask 0n10ff of each pixel in the overlapping portion can be determined in such a way that, as shown in Table 1, priority is given to the one calculated later, in principle. In this case, if the desired contour cannot be obtained, it can be corrected by changing the onloff of the cluster and then redrawing the cutting area. Note that the decision rules for this overlapping part can be changed as necessary.
For example, it is also possible to change to an OR rule as shown in Table 2 or an AND rule as shown in Table 3. In Tables 1 to 3, X
is the value of the mask that was already written, y is the value of the newly calculated mask, 2 is the newly written mask, and if the desired mask cannot be obtained by changing the onloff of the cluster, The training area can be relocated 1. For example, by relocating the training area closer to the cutting area, quality can be improved. For areas that cannot be cut out even with the above method, for example, areas where the outline does not originally appear, such as extreme highlight areas or shadow areas, manual additional corrections can be made. As a correction method, one method is to specify the position to be corrected using a position input means such as a board digitizer or a mouse, and then fill in an area of a certain size centered at that position with the mask turned on. mask off area
You can create a mask by using either of these methods, such as erasing the mask as a mask, or specifying two point positions and connecting the two points with a mask-on line segment. Upon completion of the above processing, the pattern to be cut out is surrounded by a mask. The unprocessed area outside the masked area can be filled in with the mask by filling it out. Furthermore, smoothing processing can be applied to the jaggedness of the contour line, if necessary. To output mask data, there are two methods: transferring the mask data to a plate collection device and collecting it together with other designs, transferring the mask data to a scanner and exposing it to film to create a mask plate, and converting outline data into vectors. A possible method is to obtain a mask plate by converting the outline data into data, transmitting the contour line data to a cutting plotter, and then printing the beer coat film. In the embodiment, the image data parameter space (
Although C, M, and Y spaces have been exemplified as color spaces, the spaces to be considered when implementing the present invention are not limited to this.
In addition to this space, there are also red (R), green (G), blue (B) color spaces, sets of differential values for each color plate, luminance Y, chromaticity I, etc.
, Q coordinates, hue H1 brightness V, and saturation C coordinates, which are subjected to calculations and conversions. As described above, according to the present invention, an excellent effect can be obtained in that an image to be cropped can be automatically cropped efficiently and with fewer discrimination errors. Note that if the training area is divided into clusters with a variable number of clusters, then something equivalent to multi-value conversion (quantization) into representative colors is first performed depending on the number of colors that make up the image.
After that, it can be converted into 2gi and divided into a background part and a real part. Thereby, it is possible to cut out an image in which the background part and the substance part are made of a plurality of colors. In particular, it is possible to accurately cut out objects with shadows. Furthermore, since the training area and cutting area are separate areas, the training area can be made smaller. As a result, the amount of calculation for clustering can be suppressed, so that the processing time can be known and the processing can be made more efficient. Furthermore, in clustering, data originally given in multiple dimensions can be handled as is. Therefore,
Compared to the method of clustering by treating it as a one-dimensional distribution projected onto a certain axis, there is less loss of information. This allows for highly accurate cutting.

[Brief explanation of drawings]

FIG. 1 is a front view including a partial perspective view showing the configuration of a cutting device according to the implementation side of the present invention, FIG. 2 is a block diagram showing the conceptual configuration of the device, and FIG. FIG. 4 is a flowchart showing the mask data creation procedure of the embodiment; FIG. 4 is a plan view showing the opening of training area instructions in an image; FIG. 5 is a flowchart; Line 10 showing a detailed processing procedure for clustering and showing an example of how to find the 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: calculation means for clusters, 28: cluster table storage memory, 30: binarization calculation means, 32: mask data storage memory.

Claims (2)

[Claims] (1) In a method of creating a cropping mask for performing arithmetic processing on image data and electronically cropping it, one or more training areas spanning the background and real parts of the image are specified, and the training area is By statistically analyzing the image data in the image, we create an evaluation function that indicates the possibility that a given color belongs to the background part or the substance part. For all pixels within the specified area, assign the pixel data to the evaluation function, determine whether the pixel belongs to the background part or the substance part, and determine whether the pixel belongs to the background part. A cutout mask creation method characterized by creating mask data from pixels determined to be. (2) In claim 1, when creating the evaluation function, all the pixel data in the training area is divided into two or more clusters, each divided cluster is displayed with a representative color, and the background portion is determined based on the displayed color. Alternatively, a method for creating a cutout mask is characterized in that clusters to be considered as real parts are selected, and an evaluation function indicating the possibility that an arbitrary color belongs to each cluster is created for each cluster.