JP2011248702A - Image processing device, image processing method, image processing program, and program storage medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image processing device extracting efficiently and precisely a color-coded region from a color image.SOLUTION: This image processing device extracting a region from an image has: reduced image creating means for creating one reduced image by dividing the image in a matrix state and generating a pixel from an aggregation of pixels in a divided range; clustering means for clustering color information of the reduced image in a predetermined value; color region extraction means for extracting a cluster with the same value as the region; contour extraction means for extracting a contour of the region extracted; and contour specification means for specifying a contour of the region in an original image separated by color from the contour extracted from the contour extraction means.

Description

  The present invention relates to a method for extracting a color-coded region in a color image, an image processing apparatus for performing the method, and the like.

  Various devices have been devised that perform character recognition using an image obtained by capturing a paper document with a scanner or a camera as input data. For example, when saving necessary information from paper documents such as newspapers, magazines, papers, etc., paper is bulky, so it is generally converted to image data with a scanner etc. and saved, but it was saved later In order to search for a specific document from image data, it is necessary to recognize, store, and store character information included in the page. Alternatively, by recognizing the character information included in the paper captured as image data, extracting and storing a specific character string, the specific information is registered in a portable information terminal or the like with less troublesome key input. The character recognition technology is also used for the above-mentioned purposes.

  By the way, on a color paper such as a magazine in a paper document, the background color represents a single coherent area, or the background is partially different from the background color of the entire document. Some may be emphasized.

  In such a character recognition apparatus for a color paper surface, it is desirable to divide the color paper surface for each region having the same color background as the pre-processing for character recognition, and perform the recognition processing for each region. This is because character recognition apparatuses generally generate a binary image by separating an input multivalued image into a foreground color and a background color, and perform character recognition using a portion drawn in the foreground color as a character. On the other hand, when a binary image is generated without performing the process of separating regions with different color backgrounds as described above, information that the color image is a different color background is lost and should be separated originally. This is because, despite the information, the characters are recognized as one unit and the necessary information cannot be accurately extracted.

  Or, if the color area is different for the purpose of emphasizing a part of the document, the color background and the color of the character color are often reversed, and this area is binarized by the same process as other areas This is because a binary image in which the character color and the background color are reversed results in a problem that character recognition cannot be performed.

  As a technique for separating regions having different colors and extracting regions composed of similar colors, a color space clustering method, a region expansion method, and the like are common.

  In the color space clustering method, the pixels of the input image are clustered in a color space such as RGB or YUV by a method such as K-means method or ISODATA (Iterative Self-Organizing Data Analysis Technique A) method, and can be regarded as the same color. In this method, a set of pixels is used as one area.

  The area expansion method compares the pixel color information with the neighboring pixels in the input image, and if the difference in color information is smaller than the threshold value, the same area is processed. There is no pixel that does not belong to any area. This is a method of separating regions having different colors by repeating the above.

  However, the above conventional techniques have the following problems. A color printed material generally expresses colors by arranging four-color halftone dots called CMYK on a paper surface. Since the halftone dots are very fine, individual dots cannot be identified by the human eye and appear as one color, so that various colors can be expressed by the arrangement method of the halftone dots. However, in recent years, the number of pixels of digital cameras and cameras mounted on personal digital assistants has been increasing, and when a color image is acquired from a color print with such a high-resolution device, it is the same for the human eye. Even in a color-visible region, variation may occur depending on the acquired pixel distribution. Therefore, when the color space clustering method is performed on the acquired image, pixels in the same color region are not considered to be the same color by human eyes due to pixel variations, and there is a problem that it is difficult to perform clustering with high accuracy. . Also in the area expansion method, the difference in color between adjacent pixels becomes large even in an area of the same color to the human eye, and there is a problem that the color area cannot be separated well. In addition, in either method, since processing is performed on all pixels of the entire image, there is a problem that processing time is increased for an image with a large number of pixels.

  In order to cope with such a problem, in Patent Document 1, a reduced image is created from a color original image, an area that can be regarded as the same color in the reduced image is obtained, and in the area that is regarded as the same color, the color of that area is determined. Color image processing apparatus and pattern characterized by obtaining type, number of color types, representative color, searching pixel corresponding to color extracted from reduced image in corresponding original image, and extracting region An extraction device is proposed.

  According to this apparatus, since only the color extracted in the reduced image is focused and the original image region corresponding to the region is extracted, the processing is performed more than the conventional method in which processing is performed on all the pixels of the entire original image. Time can be shortened.

Japanese Patent Laid-Open No. 11-288465

  However, in Patent Document 1, an area regarded as the same color in a reduced image is obtained, and in that area, the color type, the number of color types, and the representative color of the area are obtained. In order to obtain the number of types, the same color area obtained on the reduced image is searched to investigate all the types of colors. Further, in order to obtain a representative color, a method is described in which a frequency distribution is obtained for each color type and the mode value is used as a representative value. In order to obtain these values, all pixels in the same color region are described. It takes a lot of processing time because it is necessary to investigate and create a histogram.

  Also, in this document, the processing time is shortened by obtaining a circumscribed rectangle of the original image corresponding to the area obtained from the reduced image and searching for pixels close to the representative color in the rectangle. Certainly, if the area obtained from the reduced image is close to a quadrangle, the circumscribed rectangle of the original image has a similar shape.Therefore, there are fewer pixels that do not belong to the same color area in the search area of the original image, and the search area Can be squeezed. However, if the area obtained from the reduced image has a complex shape that is not close to a rectangle, the circumscribed rectangle of the corresponding original image has many pixels that do not belong to the same color area, so the search range for the same color area cannot be narrowed so much. Since all the pixels in the circumscribed rectangle of the original image cannot be compared with the representative colors of the area or all of the obtained color types, it must be searched whether they belong to the same area. In many cases, the problem of increased time cannot be solved. In addition, since character information is usually recorded in a region close to a quadrangle, even if a region having a complicated shape is extracted, it is often unnecessary in terms of extraction for character recognition.

  The present invention has been made in view of the above-described problems, and an object of the present invention is to provide an image that can efficiently and accurately extract a color-coded region using a reduced image with reduced color variation. It is to provide a processing apparatus.

  A further object of the present invention is to provide an image processing apparatus that can further reduce the processing time by removing an area that cannot be a character area from the extracted area.

  In order to solve the above problems, an image processing apparatus according to the present invention is an image processing apparatus that extracts an area from an image, divides the image into a matrix, and generates one pixel from a set of pixels in the divided range. A reduced image generating means for generating a reduced image, a clustering means for clustering the color information of the reduced image into a predetermined value, a color area extracting means for extracting clusters having the same value as an area, It is characterized by comprising contour extracting means for extracting a contour and contour specifying means for specifying the contour of the region of the original image color-coded from the contour extracted by the contour extracting means.

  Further, the image processing apparatus according to the present invention is characterized in that the reduced image generation means uses an average value of a set of pixels in the divided range as pixels of the reduced image.

  The image processing apparatus according to the present invention is characterized by comprising cluster expansion / contraction means for expanding or contracting the connection of pixels belonging to the same cluster.

  Further, the image processing apparatus according to the present invention is characterized in that the color area extracting means does not extract an area where the number of pixels in an area where pixels belonging to the same cluster are connected is smaller than a predetermined value.

  In addition, the image processing apparatus according to the present invention is characterized in that the color region extraction unit evaluates the shape of a region in which pixels belonging to the same cluster are connected, and does not extract a region that does not match the evaluation value.

The image processing method according to the present invention is an image processing method for extracting a region from an image. The image is divided into a matrix, and a reduced image is generated by generating one pixel from a set of pixels in the divided range. A reduced image generating step, a clustering step for clustering color information of the reduced image to a predetermined value, a color region extracting step for extracting a cluster having the same value as a region, and a contour extracting step for extracting the contour of the extracted region And an outline specifying step for specifying the outline of the color-coded original image area from the outline extracted by the outline extracting step.

  The image processing program according to the present invention is an image processing program that extracts an area from an image, divides the image into a matrix, generates one pixel from a set of pixels in the divided range, and generates a reduced image. A reduced image generating step, a clustering step for clustering color information of the reduced image to a predetermined value, a color region extracting step for extracting a cluster having the same value as a region, and a contour extracting step for extracting the contour of the extracted region And a contour specifying step for specifying the contour of the color-coded region of the original image from the contour extracted by the contour extracting step.

  A storage medium according to the present invention stores the image processing program.

  According to the image processing apparatus of the present invention, it is possible to efficiently and accurately extract a color-coded region using a reduced image with reduced color variation, and further, the extracted region can be a character region. The processing time can be shortened by removing such areas.

It is a block diagram which shows the structure of the image processing apparatus which concerns on embodiment. It is a flowchart which shows the flow of the process which concerns on embodiment. It is a cluster map concerning an embodiment. It is an example of a cluster map with pixel noise. It is a cluster map concerning an embodiment. It is a cluster map concerning an embodiment. It is a cluster map concerning an embodiment. It is a color area map concerning an embodiment. It is explanatory drawing about the contour extraction process which concerns on embodiment. It is explanatory drawing about the contour extraction process which concerns on embodiment. This is an example in which the outline of the color region is complicated. It is a figure which shows the color area | region in the original image which concerns on embodiment. It is the figure which matched the color area extracted in the reduction image which concerns on embodiment with the original image. It is a figure showing the direction of the outline of the color area which concerns on embodiment. It is the figure which showed the range in which the extraction range of the outline which concerns on embodiment was set. It is the figure which showed an example of the edge direction table which concerns on embodiment. It is a figure which shows the method of Prewitt in an outline extraction process.

  Hereinafter, with reference to the drawings, an example of extracting color-coded character areas from color paper will be described with respect to an embodiment of an image processing apparatus according to the present invention. In the drawings of the present invention, the same reference numerals represent the same or corresponding parts.

  FIG. 1 is a block diagram showing a configuration of an image processing apparatus for carrying out the present invention. An image processing apparatus according to the present invention includes an image input unit 101 that inputs a color image, an image memory 102 that stores the input color image, and a grayscale image generation that generates grayscale image data from the color image stored in the image memory 102. Unit 103, a grayscale image memory 104 that stores the generated grayscale image, a reduced image generation unit 105 that generates reduced image data from a color image stored in the image memory 102, and a reduced image that stores the generated reduced image Memory 106, clustering unit 107 that clusters color information of reduced images, cluster expansion / contraction unit 108 that expands / contracts connection of pixels belonging to the same cluster, and color region extraction unit that extracts connection of pixels belonging to the same cluster as a region 109, clustering classification and extraction of color information A color area memory 110 that stores color areas, a color area deletion unit 112 that deletes a color area that matches a predetermined condition with respect to the color areas stored in the color area memory 110, and tracks the outline of the extracted color area A contour extraction unit 111 that extracts a contour pixel, a contour specification range setting unit 113 that sets a range of a color image corresponding to the contour pixel, an edge direction table 114 that stores an edge direction of the contour pixel, and a contour specification of the set color image It is comprised from the outline specific | specification part which specifies the outline of the color area | region of a color image from a range.

  FIG. 2 is a flowchart showing the overall processing flow of the present invention. First, the image processing apparatus is activated, and the image input unit 101 converts the color paper surface into electronic data such as a CCD camera or a scanner and inputs it, and stores it in the image memory 102 as a color image (hereinafter referred to as an original image). (Step S201). In this embodiment, a case where color pixels are expressed by R, G, and B will be described. However, other color systems such as YUV and HSV may be used.

  Next, the gray scale image generation unit 103 generates a gray scale image from the original image stored in the image memory 102 and stores it in the gray scale image memory 104 (step S202). As a process for generating a gray scale image, there is a method of calculating a luminance value by a weighted average of R, G, and B to obtain an image value.

  Subsequently, the reduced image generation unit 105 generates a reduced image from the original image stored in the image memory 102 and stores it in the reduced image memory 106 (step S203). For example, the following method can be used to generate a reduced image.

  In order to generate a reduced image in which N × N pixels of the original image stored in the image memory 102 are one pixel, the original image is divided into a matrix for each N × N pixel and is within each divided range. For the pixel values, an average value is calculated for each of R, G, and B, and one pixel is assigned according to the average values of R, G, and B. Here, when N is too small, the total number of pixels of the reduced image increases, so it takes time in the subsequent processing, and when the halftone dot of the color print is rough, the variation in the pixels of the reduced image increases. In the clustering described later, the accuracy also deteriorates. If N is too large, it is difficult to extract a small color area. In general, N is about 4 to 16, and it is desirable to select a value according to the input target image. By using a reduced image in this way, it is possible to extract a color-coded region efficiently. In addition, by using the average value of the pixels of the original image as a reduced image, it is possible to extract a color-coded region with high accuracy using a reduced image in which the color variation in the original image is reduced.

  Next, the clustering unit 107 extracts color information of the reduced image stored in the reduced image memory 106 and performs clustering processing. Clustering is to find a group of clusters (cluster) in some feature space in a data set. In the present invention, clustering processing is performed by clustering the color space, the pixels on the reduced image are classified into clusters, a cluster map indicating the distribution status is created, and stored in the color area memory 110 (step S204). As a specific method of color space clustering, an existing method such as a K-means method or an ISODATA method may be used.

  FIG. 3 shows an example of a cluster map, which is classified into four-color clusters of numbers 0 to 3. Since the purpose of the present invention is to extract a color region different from the entire image, the subsequent processing is performed for clusters other than the background color. When a general paper image is captured, the background color is often white, but various colors can be considered depending on the type of color print, so the number of pixels belonging to the largest number in the classified cluster You can choose. In the example of FIG. 3, it is determined that the number 0 is a cluster representing the background color.

  Next, the cluster expansion / contraction unit 108 performs expansion processing and contraction processing for each cluster on the cluster map (step S205). The expansion processing is processing for converting the value of the pixel in the background component in contact with the boundary pixel of a certain graphic component into the value of the graphic component pixel and expanding it by one pixel. Conversely, the contraction process is a process of converting all the pixel values at the boundary of a certain graphic component into pixels of the background component and reducing it by one pixel. These processes are performed for the purpose of removing noise such as missing pixels or unnecessary pixels. In this embodiment, expanding a set of pixels belonging to a target cluster is referred to as expansion processing, and contracting is referred to as contraction processing.

  FIG. 4 is an example of a cluster map with pixel noise. In this example, there are two unnecessary isolated pixels 10 in the zeroth cluster due to noise, and the second cluster has a pixel defect 11 partly in a shape that should originally be a rectangle. In the color area extraction process in the next step S206, which will be described later, a labeling process (a process of adding the same label to connected pixels) is performed for each cluster in the cluster map, and the connected area is extracted to be a color area. For this reason, a part of the region may be lost or an unnecessary small region may be extracted. Therefore, the cluster map is subjected to expansion processing and contraction processing for each individual cluster to remove noise.

  FIG. 5 is a cluster map after the expansion process is performed on the cluster map of FIG. In the cluster map of FIG. 4, expansion processing is performed in the vicinity of 8 (periphery in the vertical and horizontal directions) for each cluster.

  FIG. 6 is a cluster map after the contraction process is performed on the cluster map of FIG. After performing the dilation process on the cluster map of FIG. 4 and further performing the contraction process in the vicinity of 8 in this state, a cluster map as shown in FIG. 6 is created and exists as a pixel defect 11 in FIG. The missing portion of the second cluster that has been filled is filled, and the original rectangular region can be extracted.

  FIG. 7 is a cluster map after the expansion process is performed after the contraction process is performed on the cluster map of FIG. By such processing, two unnecessary areas of the first cluster that existed as isolated pixels 10 in FIG. 4 are removed.

  By performing the expansion / contraction process in this way, even when the cluster distribution lacks or there are extra pixels at the clustering stage, it is possible to extract the color-coded region with high accuracy.

  Next, the color area extraction unit 109 performs a labeling process for each cluster in the cluster map stored in the color area memory 110, extracts a connected area, creates a color area map, and stores the color area map in the color area memory 110. (Step S206). At that time, in order to exclude a small area that cannot be a color-coded character area, the number of pixels constituting the connected area is counted for the extracted connected area, and the number of pixels is larger than a predetermined threshold. Only the color area memory 110 is stored. The threshold value may be set as appropriate according to the reduced image. In this way, the processing time can be shortened by not extracting a small area that cannot be a color-coded character area from the extracted color area.

  FIG. 8 shows a result of creating a color area map by performing a labeling process on clusters other than the background color with respect to the cluster classification of FIG. In this example, the clusters of R2 and R3 are connected, but labeling is performed as another label even if regions having different clusters are connected. Then, unique numbers R 1 to R 4 are assigned to these labels to create a color area map and store it in the color area memory 110.

  The next processing from step S207 to S210 is performed for each color region extracted in step S206.

  First, the contour extraction unit 111 performs a contour extraction process in order to extract a pixel that is a contour of a color region in a reduced image, using the color region map stored in the color region memory 110 (step S207). The contour extraction process is a process of extracting a contour line so as to follow the boundary line of the region.

  FIG. 9 is an explanatory diagram of a 4-connected contour extraction process used in the present embodiment. First, as shown in FIG. 9A, raster scanning is performed from the upper left of the image as indicated by a dashed arrow, and the first pixel found in the region where the contour is to be extracted (the pixel marked with ●) is set as the starting point. Since it has been confirmed by the raster scan that there is no pixel at the top and left at the start point position, it is checked whether there is a pixel in the extraction direction in the order of bottom and right. If there are no pixels in either direction, the process ends as an isolated point. In this figure, since there is a pixel below, the process proceeds to a lower point as shown in FIG. 9B, and this point is set as a new extraction point. Thereafter, it is checked whether or not there is a pixel with a counterclockwise priority from the pixel on the right hand side with respect to the traveling direction of the extraction point. This priority order is represented by 1 to 4 in the figure. In FIG. 9B, since the traveling direction is downward, as shown in the figure, it is checked whether there is a pixel in the order of left, bottom, right, and top, and proceeds in that direction when there is a pixel. In this case, there is no pixel on the left and there is a pixel below, so the process proceeds downward.

  Similarly, the contour extraction process proceeds, and in FIG. 9C, since the traveling direction is right, as shown in the figure, it is checked whether there are pixels in the order of bottom, right, top, and left. In this case, since there is a pixel below, the pixel below is set as a new extraction point. The above process is repeated, and when the extraction point returns to the start point as shown in FIG.

  FIG. 10 is a diagram obtained by performing the above contour extraction processing on the color region R1 extracted in FIG. The contour line extracted by this contour extraction processing is stored in the color area map of the color area memory 110. In the contour extraction process, it can be seen that there is a contour line in the direction advanced when the contour is extracted, that is, in the direction of the arrow in FIG. 10, so the direction of the contour line is stored together.

  At this time, the length of the extracted outline is counted. In the contour line length counting method, the number of pixels traced from the start pixel to the start pixel again is defined as the contour line length. For example, in FIG. 9, the start pixel is 3 downward, 3 right, 1 downward, 9 right, 3 upward, 3 left, 1 upward, and 9 left. To return to the start pixel. Therefore, the total length of the contour line is 32.

Next, in the color area erasing unit 112, an area having a complicated shape that cannot be a color-coded character area is removed from the extracted color area (step S208). Specifically, the evaluation value is calculated using the calculation formula “(length of contour line) ² / number of constituent pixels” using the length of the contour line and the number of pixels constituting the color region. This value becomes larger as the contour of the color region is complicated and the shape becomes complicated.

  FIG. 11 shows an example in which the outline of the color region is complicated. In the color region shown in FIG. 11, the number of constituent pixels is 54, which is the same as that shown in FIG. However, when the contour tracking process is performed, the length of the contour line is 62, which is longer than the length 32 of the contour line in FIG. When the evaluation values are calculated for the color regions shown in these two figures, it is found that FIG. 9 is 19.0 and FIG. 11 is 71.2, and FIG. 11 is considerably larger. If the evaluation value is equal to or greater than a predetermined threshold value, the evaluated color area is deleted from the color area map stored in the color area memory 110. As this threshold value, a value for determining that the shape of the color region is complicated may be appropriately set according to the reduced image. In this way, the processing time can be shortened by removing a region having a complicated shape that cannot be a character region color-coded from the extracted color region.

  Next, the contour specific range setting unit 113 refers to the color region map and sets the contour extraction range on the original image (step S209). Here, since the reduced image is created with an N × N range of pixels of the original image as one pixel, if the coordinates of the reduced image are {m, n}, the corresponding original image is {m * N, n * N} is in the upper left, and {(m + 1) * N−1, (n + 1) * N−1} is in the lower right range. Here, since the reduced image uses the average value of the color information in the N × N range of the original image as the pixel value, the color area of the original image does not necessarily fall within the range of the original image corresponding to the outline pixel on the reduced image. It does not include the outline. This will be described with reference to FIG.

  FIG. 12 shows color regions in the original image. Blocks divided by broken lines indicate ranges corresponding to individual pixels of the reduced image.

  FIG. 13 shows a diagram in which a color area extracted in a reduced image is associated with an original image, and a block with “x” indicates that an outline of the extracted color area exists. In this way, when the block having the outline of the color area includes a large number of color area portions in the original image when the reduced image is generated, the color information of the pixel of the reduced image calculated by the average value is completely colored. Since it is close to the color information of the pixel of the reduced image including the area, it is classified into the same cluster and extracted as the same color area. On the other hand, when the color area portion in the original image is not included so much, the color information of the pixel is not completely close to the color information of the pixel including the color area, and thus is not classified into the same cluster. However, in that case, the outline of the original image exists in the outer block adjacent to the outline of the extracted color area. Therefore, the contour extraction range is expanded depending on the direction of the extracted contour.

  The arrow in FIG. 14 represents the direction of the outline of the color region shown in FIG. In this way, the block outside the contour line becomes a block on the right hand side with respect to the direction of the contour line. For example, in the {3, 3} block, the direction of the contour line is left, so that the contour extraction range is extended to the upward block as shown in the figure. Further, in the {2, 3} block, since it hits the corner of the color region, there are contour lines in the left and upper directions. In this case, as shown in the figure, the contour extraction range is extended to a square block including the top and the left. The above processing is performed for all the blocks where the contour exists.

  FIG. 15 shows a range in which a contour extraction range is set. Furthermore, the corresponding edge direction table 114 is created.

  FIG. 16 shows an example of the edge direction table 114. The horizontal is the X coordinate of the reduced image, the vertical is the Y coordinate of the reduced image, and the arrow in the table is the direction of the edge to be extracted. Here, the edge direction is the direction from the extracted color area to the background area, that is, the direction on the right hand side with respect to the direction of the contour line. For blocks having contour lines in a plurality of directions, a plurality of edge directions are set as values.

  Next, the contour specifying unit 115 performs a contour specifying process within the range of the original image set in step S209 (step S210). The contour specifying process is a process for detecting an edge as a contour using various methods such as a differential operator and template matching, and is a generally known image processing technique. In this embodiment, description will be made using a Prewitt method which is a typical template matching.

  FIG. 17 shows the edge direction and the mask pattern in the Prewitt method in the contour extraction process. In the Prewitt method, as shown in FIG. 17, eight types of mask patterns corresponding to eight edge directions are prepared as standard patterns representing contours. The input pixel and its neighboring pixels are summed by multiplying the input pixel at the center of the mask pattern by the value of the mask pattern, and the calculated value becomes the strength in the direction of the edge corresponding to the mask pattern. As the input image, a grayscale image stored in the grayscale image memory 104 is used.

  The contour specifying unit 115 refers to the edge direction table 114 and calculates the strength of the edge using the corresponding mask pattern within the range of the original image for which a value is set in the table. For example, in {1, 2} of the edge direction table 114 of FIG. 16, it can be seen that there are left and upward edges. The corresponding range of the original image is a range with {N, 2N} at the upper left and {2N-1, 3N-1} at the lower right. Using the templates (b) and (h) and the surrounding templates (a), (c), and (g) in FIG. 17 for calculating the edge in the direction, the direction in which the edge is detected and the strength of the edge are determined. calculate. Similarly, since {3, 2} has an upward edge, the range of the corresponding original image is a range in which {3N, 2N} is upper left and {4N-1, 3N-1} is lower right. Using the template (b) for calculating the edge in the upward direction and the surrounding templates (a) and (c) for the pixels of the grayscale image in the range, the direction in which the edge is detected most Calculate the edge strength.

  Such processing is performed for all the ranges for which values are set in the edge direction table 114, and the outline of the color region in the original image is specified. In a general contour specifying process, the direction of the edge detected in each pixel is not known, so a template corresponding to all directions must be applied to each pixel. In this embodiment, the edge direction table 114 is used. Since the direction of the contour line is stored, only the template corresponding to the direction of the contour line needs to be used, and the edge detection processing time can be greatly shortened.

  Next, it is checked whether or not there is a color region that has not been subjected to contour identification processing (step S211). If there is, the process returns to step S207, and the processes from step S207 to S210 are repeated for the color region that has not yet undergone the contour specifying process.

  According to the present embodiment, it is possible to generate a reduced image in which the color variation is smaller than that of the original image, and to extract the color-coded region efficiently and accurately using the reduced image. The processing time can be shortened by removing an area that cannot be a character area.

  The present invention is not limited to the embodiment described above, and various modifications can be made within the scope indicated in the claims. The processing target is not limited to the color paper surface, and the distance image to the object is color-coded from the distance image. It can be applied to various color images such as extraction of color areas.

DESCRIPTION OF SYMBOLS 101 Image input part 102 Image memory 103 Gray scale image generation part 104 Gray scale image memory 105 Reduction image generation part 106 Reduction image memory 107 Clustering part 108 Color area extraction part 109 Cluster expansion / contraction part 110 Color area memory 111 Outline extraction part 112 Color Area erasing unit 113 Outline specifying range setting unit 114 Edge direction table 115 Outline specifying unit

Claims (8)

  In the image processing apparatus for extracting an area from an image, reduced image generating means for dividing the image into a matrix and generating one pixel from a set of pixels in the divided range to generate a reduced image, and the color of the reduced image Clustering means for clustering information into predetermined values, color area extracting means for extracting clusters having the same value as areas, outline extracting means for extracting outlines of the extracted areas, and contours extracted by the outline extracting means And an outline specifying means for specifying the outline of the area of the color-coded original image.   The image processing apparatus according to claim 1, wherein the reduced image generation unit uses an average value of a set of pixels in a divided range as a pixel of the reduced image.   The image processing apparatus according to claim 1, further comprising a cluster expansion / contraction unit that expands / contracts connections of pixels belonging to the same cluster.   4. The image processing apparatus according to claim 1, wherein the color area extraction unit does not extract an area in which the number of pixels in an area where pixels belonging to the same cluster are connected is smaller than a predetermined value. 5. The image processing apparatus described.   5. The image processing apparatus according to claim 1, wherein the color region extraction unit evaluates the shape of a region in which pixels belonging to the same cluster are connected, and does not extract a region that does not match the evaluation value. An image processing apparatus according to claim 1.   In the image processing method for extracting an area from an image, a reduced image generating step of dividing the image into a matrix and generating one pixel from a set of pixels in the divided range to generate a reduced image, and the color of the reduced image A clustering step for clustering information into predetermined values, a color region extraction step for extracting clusters having the same value as a region, a contour extraction step for extracting a contour of the extracted region, and a contour extracted by the contour extraction step And an outline specifying step for specifying the outline of the area of the color-coded original image.   In the image processing program for extracting an area from an image, a reduced image generation step of dividing the image into a matrix and generating one pixel from a set of pixels in the divided range to generate a reduced image, and the color of the reduced image A clustering step for clustering information into predetermined values, a color region extraction step for extracting clusters having the same value as a region, a contour extraction step for extracting a contour of the extracted region, and a contour extracted by the contour extraction step An image processing program for causing a computer to execute an outline specifying step for specifying the outline of a color-coded original image area.   A computer-readable recording medium on which the image processing program according to claim 7 is recorded.