JP3647075B2 - Image search method and apparatus - Google Patents

Description

[0001]
[Industrial application fields]
The present invention relates to an image search method and apparatus for searching for image data similar to an image or a graphic specified by an image database or an image file system, for example.
[0002]
[Prior art]
Conventionally, the following method has been used as an image search method for searching for a desired image from a plurality of images stored in an image file format.
(1) When an image to be searched is input, the image and the image data stored in the image file are sequentially superimposed, and the difference in pixel between the search image and each image in the image file is determined. A method of obtaining the sum and using the so-called template matching method in which the minimum value is the difference between the search image and the image data of the image file, and selecting image data with this difference equal to or less than a predetermined value as a similar image (details of the template matching method are , Textbooks on image processing, for example, introduction to computer image processing, Soken Publishing, 5.6 Matching, pp.148-152).
(2) The image data stored in the image file is divided into several uniform regions in advance, and the correspondence between the data describing each image region and the search image or the search figure is expressed by area · 2 A method of calculating using similarity such as second moment and selecting a similar image having a large similarity (for example, image search using a schematic image: Junji Hirata, Yoshinori Hara, Information Processing Society of Japan, DE92, DE92) -2).
[0003]
[Problems to be solved by the invention]
However, in the conventional example (1), it is necessary to move the designated image with respect to the image data to be searched, and to calculate the difference in the number of pixels corresponding to each movement. Was. In addition, since an accurate search result cannot be obtained unless the image instructed for the search is moved and compared at a sufficiently small distance within the two-dimensional region of the image data, a very large amount of calculation time is required. There was a problem.
[0004]
Further, in the above conventional example (2), there is a problem that if one area of the image data of the image file to be searched does not directly correspond to the image instructed as a search target, it cannot be correctly searched. .
[0005]
Here, in the so-called region division method that automatically divides image data into a plurality of regions, the target area that humans recognize as one region is often divided into a plurality of parts and extracted. This limitation has been a major practical problem.
[0006]
The present invention has been made in view of the above conventional example, and provides an image search method and apparatus capable of efficiently searching image data even when a search area in image data is divided into a plurality of areas and extracted. The purpose is to do.
[0007]
It is another object of the present invention to provide an image search method and apparatus capable of searching for an image based on image data divided in accordance with an area of image data recognized as one area.
[0008]
Another object of the present invention is to provide an image search method and apparatus capable of efficiently searching for an image by reducing the time required to calculate the similarity of images.
[0009]
[Means for Solving the Problems]
In order to achieve the above object, the image search apparatus of the present invention has the following configuration. That is,
An image search apparatus for searching image data including an image similar to a search-designated image using image description information describing information for each region in the image as reference information, wherein each of the regions in the image A combination of each of the single regions and the single region processing means for detecting a single region included in the search image from the individual regions in accordance with each and a search image designated for search, and the search A composite area processing means for obtaining a composite area obtained by combining the single areas included in the search image according to an image; a similarity detection means for obtaining a similarity between the composite area and the search image; and the similarity Search means for searching for image data in accordance with the similarity determined by the detection means.
[0010]
  In order to achieve the above object, the image search method of the present invention includes the following steps. That is, search specifiedSearchImage data including images similar to imagesFrom image fileSearch forIn image retrieval deviceAn image search method,
  Storing a search image instructed to be searched in a search image memory;
  With reference to a description file describing information about the position, shape and color information of individual regions in the image stored in the image file,Individual areas in the imageStored in the search image memoryIncluded in search imageA determination step of determining whether or not,
  Storing the description information of the area determined to be included in the search image in the determination step, and the movable area of the search image with respect to the image in a single area memory;
  Based on the description information and the movable area,Said single regionWhen it is determined that each area stored in the memory satisfies the conditions included in the search image at the same timeFinding a composite area that combines the above areasTo store in multiple area memoryProcess,
  SaidStored in the multi-region memoryThe similarity between the composite area and the search imageCalculation to calculateProcess,
  SaidCalculated in the calculation processDepending on similarityContains the search imageAnd a step of retrieving image data.
[0011]
[Action]
In the above configuration, a single region included in the search image is detected from the individual regions in accordance with each of the individual regions in the image and the search image in which the search is designated, and each of the single regions is detected. In accordance with the combination and the search image, a composite area obtained by combining the single areas included in the search image is obtained. Then, the similarity between the composite area and the search image is obtained, and the image data is searched according to the obtained similarity.
[0012]
【Example】
Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.
[0013]
FIG. 1 is a block diagram functionally showing the flow of processing in an image search apparatus according to an embodiment of the present invention.
[0014]
In FIG. 1, reference numeral 100 denotes an image to be searched (hereinafter referred to as a search image), and the search image 100 instructed to search is stored in the search image memory 101. A single area processing unit 102 refers to the description file 110 in which the contents of each image data are described using the characteristics of the image data, and whether each area included in each image data is included in the search image. If it is included, the description information of the area of the image data and the information of the movable area of the search image are added to the single area memory 103. A composite area processing unit 104 determines whether or not a composite area satisfies a condition that all areas constituting the composite area are included in the search image at the same time. If included, the description information of the area is included. And the information on the movable area of the search image are added to the composite area memory 105.
[0015]
A maximum value selection unit 106 calculates the similarity between the search area and the composite area stored in the composite area memory 105 by the composite area processing unit 104, and calculates the maximum value of the similarity value for each image area. Select as similarity. Reference numeral 107 denotes an ordering unit for inputting the similarity of the search image with respect to each image region, and selecting K pieces of image data from the higher similarity. An image data access unit 109 accesses the image file 109 and reads out the image data selected by the ordering unit 107. Reference numeral 111 denotes a display unit that displays the searched images.
[0016]
The single area processing unit 102 includes a search image memory 101 storing a search image (including graphics) given by a searcher who is a user, and an area of each area in the image data stored in the image file 109. A reference is made to an image description file 110 in which a boundary point coordinate sequence and the like are described, and each region of the image description file 110 is a region that can be moved under the condition that the region is included in the search image. A possible area is calculated, and it is determined whether or not the condition that the area is included in the search image is satisfied. If included, the description information of the area and the information on the movable area of the search image are simply set. It is added to the one area memory 103.
[0017]
The composite area processing unit 104 inputs the contents (search image) of the search image memory 101, the area information in the single area memory 103, and the composite area information in the composite area memory 105, and combines the single areas. When a composite area and a composite area combining a composite area and a single area are generated, whether or not the condition that these composite areas are included in the search image is satisfied, and the area is included in the search image Then, the description information of the composite area and the information of the movable area of the search image are added to the composite area memory 105.
[0018]
The maximum value selection unit 106 selects the composite area and the search image for the composite area in the composite area memory 105 that satisfies the condition that the composite area is included in the search image, selected by the composite area processing unit 104. The similarity is calculated, the maximum value of the similarity for all the regions included in one image data is selected, and this is output as the similarity to the image data.
[0019]
The ordering unit 107 inputs the similarity to each image data obtained from the maximum value selection unit 106, and selects k pieces of image data in descending order of similarity. The data access unit 108 uses the similarity information of each sorted image data stored in the internal memory of the ordering unit 107, and outputs a plurality of corresponding image data from the image file 109 in descending order of similarity. The data is read out and displayed on the display unit (display) 111.
[0020]
FIG. 2 is a diagram illustrating a display example of a screen in the image search apparatus according to the present embodiment.
[0021]
In FIG. 2, reference numeral 201 denotes a graphic input window for writing an image (search image) that the user wants to search. Reference numeral 202 denotes a condition input window for the user to specify the type of search image. Reference numeral 203 denotes a search result display window for displaying images of search results in the order of similarity, for example. A color palette 204 is used to select color information to be given to the search image by the user. Reference numeral 205 denotes an attached information display window for displaying attached information attached to an image as a search result.
[0022]
In the above configuration, the user designates and inputs the type of image data (search image) to be searched for in the condition input window 202, and a graphic imitating a partial image in the image data to be searched is displayed in the graphic input window 201. The color of the figure is designated using the color palette 204.
[0023]
As a result of searching the image data based on the specified image search condition, image data indicating the search result is displayed in the search result display window 203, and information accompanying the search result image is attached. It is displayed on the information display window 205.
[0024]
FIG. 3 is a diagram for explaining image search processing according to the first embodiment of the present invention, and is a functional configuration diagram showing processing for realizing the single area processing unit 102 shown in FIG.
[0025]
FIG. 4 is a diagram illustrating an example of a search image. Here, the search image is a polygonal figure, and the search image information includes, for example, a specified figure vertex row, figure vertical and horizontal width, figure area, color information, and the like. A region serving as a unit of region description information is obtained by applying region division processing known to a technician in the relevant field to search target image data stored in the image file 109. Such region segmentation processing is described in, for example, Tominaga Shoji: “Color classification and segmentation of color images”, Journal of Information Processing Society of Japan, Vol. 31, no. 11, pp. 1589-1598, 1990, Nov. Is shown in
[0026]
FIG. 5 is a diagram schematically illustrating an example of a region division result of image data to be searched. Here, the region description information includes (region outline point sequence, region starting point, region length and width, region area, region average color information, etc.) for each region.
[0027]
3, 301 is a color information determination unit, 302 is a size information determination unit, 303 is a movable region calculation unit, and 304 is a region information determination unit.
[0028]
Based on the input search image and the region description information of the image data to be searched, the color information determination unit 301 calculates the difference between the two color information, and if the difference is greater than a predetermined threshold, there is a mismatch. It is determined that it will not be part of the search image, and is excluded from the search target image. This difference in color information can be calculated as follows when the color of the search image is (R, G, B) and the color of the area of the image data is (r, g, b).
[0029]
D = | R−r | + | G−g | + | B−b |
Next, the size information determination unit 302 compares the vertical width and the horizontal width of the two. For example, if the vertical or horizontal width of each divided area shown in FIG. It is excluded from the search target.
[0030]
If the color information determination unit 301 and the size information determination unit 302 do not determine that the images do not match, the movable region calculation unit 303 adds the search image to each region of the region description information of the search target image. A movable area indicating a parallel movement range including the area is calculated. Then, the area information determination unit 304 determines whether or not the calculated movable area is an empty area. If the movable area is empty, the area information determination unit 304 determines that there is a mismatch and excludes the image data from the search target. . In this way, only the region that can be included in the search image is selected as the search target image.
[0031]
FIG. 6 is a flowchart showing detection processing of a boundary line that is a boundary of the movable region in the movable region calculation unit 303 of FIG.
[0032]
First, in step S1, the maximum value is set to “−9999999”, and the formula of the line segment F is set to F = aj × x + bj × y. Next, in step S2, the equation of the edge line Ej connecting the adjacent vertex coordinates Qj and Qj + 1 from the vertex row Qj (j is an index corresponding to each vertex) of the search image.
aj × x + bj × y + cj = 0
From the coefficients (aj, bj) of the equation of the edge Ej and the coordinates (Xi, Yi) of the boundary point Pi for the boundary point sequence Pi (i is an index corresponding to each point) of each region
F = aj * Xi + bj * Yi
Is calculated (step S3). In step S4, the maximum value FMAX (j) of F with respect to the boundary point sequence of one region is calculated (note that the order of the vertex sequence of the search image needs to be arranged in the clockwise direction. If is counterclockwise, rearrange in reverse order in advance). When this maximum value is larger than the value obtained so far, the maximum value is updated in step S5. In step S6, it is checked whether or not the processing for all the boundary points P has been completed. If not, the process proceeds to step S7, the index i is incremented by 1, and the process returns to step S2.
[0033]
When FMAX (j) that defines the boundary line aj × x + bj × y−FMAX (j) = 0 including all the points in the region boundary point sequence Pi is obtained in this way, the FMAX is registered in step S8. In step S9, it is checked whether or not this processing has been performed for all the ridge lines Ej composed of the vertex sequence of the search image. Find all boundaries.
[0034]
FIG. 7 is a schematic diagram for explaining the processing result shown in the flowchart of FIG. Here, when the search image (FIG. 4) is translated, an area 700 in the center of the figure indicates a limit included in the search image by an arrow. The straight line indicated by the tip of these arrows is represented by an equation.
aj * x + bj * y-FMAX (j) = 0
It is.
[0035]
For the region 700 in the image in FIG. 7, the boundary point sequence Pi = (Xi, Yi) of the region 700 is
aj × Xi + bj × Yi <= FMAX (j)
And a geometric parameter space with the X and Y translation amounts (dx, dy) as axes is considered, the equation corresponding to each ridge line is
From aj * (x-dx) + bj * (y-dy) + cj = 0
aj * Xi + bj * Yi-FMAX (j) <= aj * (Xi-dx) + bj * (Yi-dy) + cj, where (dx, dy) must satisfy the following conditions.
[0036]
aj * dx + bj * dy- (cj + FMAX (j)) <= 0
From this, the boundary line in (dx, dy) space
aj * dx + bj * dy- (cj + FMAX (j)) = 0
By drawing, the existence range of (dx, dy) that satisfies the conditions for all the edge lines of the search image is obtained as shown in FIG. 8, for example. If the existence range of the point (dx, dy) that satisfies the conditions for all the edge lines of the search image is empty, it is determined that the area of the image data to be searched does not fit in the search image.
[0037]
FIG. 9 is a flowchart showing a determination process of whether or not the movable area is an empty area in the area information determination unit 304 of FIG.
[0038]
In step S21, the positions of the intersections between the boundary lines in the (dx, dy) space are calculated. Determine whether. Then, if any of the intersections between the boundary lines satisfies the conditions for all the ridgelines of the search image, it is determined that the movable area is not empty. Here, the position of the intersection between the boundary lines is calculated using the simultaneous equation solving method for the equations of the two boundary lines.
[0039]
FIG. 10 is a flowchart showing the flow of processing for determining the presence or absence of an internal intersection in step S22 of FIG.
[0040]
In step S31, the intersection Pi = (Xi, Yi) between the boundary lines is set, and the boundary line Bj: aj * dx + bj * dy- (cj * FMAX (j)) = 0 with respect to the set intersection Pi. Conditions for
aj * Xi + bj * Yi <= cj + FMAX (j)
Is satisfied for all the boundary lines in order (step S33), it is checked whether the processing for all the boundary lines is completed in step S34, and if all the boundary lines are satisfied, Then, it is determined that there is an internal intersection and the process is terminated. On the other hand, when there are unsatisfied conditions, the above determination is repeated for the next intersection.
[0041]
In step S36, it is determined whether the determination process for all intersections has been completed. If it is not determined that there is an internal intersection even if all the intersections are determined, it is determined that there is no internal intersection, and the process ends.
[0042]
As a result of such determination, for example, several areas as schematically shown in FIG. 11 are inserted into the single area memory 103.
[0043]
FIG. 12 is a flowchart showing the flow of composite area generation and processing in the composite area processing unit 104 of FIG. In this process, a composite area combining the area information in the single area memory 103 is generated and inserted into the composite area memory 105.
[0044]
In step S41, the system first sets all areas in the single area memory 103 in the complex area memory 105 as complex areas. In step S42, the single area i in the single area memory 103 is sequentially selected. In step S43, the temporary memory for temporarily storing the composite area generated by the combination of the selected single area i and the composite area (including the copied single area) of the composite area memory 105 is cleared. .
[0045]
In step S44, the composite area j (including the copied single area) in the composite area memory 105 is selected in order. In step S45, it is checked whether or not there is a selected single region i among the single regions constituting the selected composite region j. If there is no single region i, the process proceeds to step S46 to generate a composite region k by combining these. . In step S47, it is determined whether or not the composite area k is included in the search image. If included, the process proceeds to step S48, and the composite area k is added to the temporary memory. Until this is performed for all the composite areas in the composite area memory 105 in step S49, the above-described processing is executed while incrementing the index j.
[0046]
When the processing for all the composite areas is completed in step S49, the process proceeds to step S51 to add the composite area in the temporary memory to the composite area memory 105, and in step S52, the process for all areas in the single area. In step S53, the above processing is repeated while incrementing the index i.
[0047]
FIG. 13 is a functional block diagram illustrating a functional configuration of the composite area processing unit 104 that performs a process of determining whether or not each composite area k is included in the search image.
[0048]
First, the size information determination unit 1301 determines the size information of the composite area. The area of the composite region k is the sum of the areas of the single region i and the composite region j included therein, and the starting point of the composite region k and its vertical and horizontal width are calculated from the starting point of each region, and the starting point and the vertical and horizontal width. The vertical and horizontal widths are calculated as the difference from the maximum value starting from the minimum value among the end points.
[0049]
If the area of the composite area k calculated in this way is larger than the area of the search image, or if the vertical width or horizontal width of the composite area k is larger than the vertical width / width of the search image, these are the search images. Are excluded because they cannot be included.
[0050]
Next, in order to determine whether the single region i and the composite region j constituting the composite region k are present in the search image at the same time, the intersection region calculation unit 1032 determines whether the single region i and the composite region j The intersection area of the movable area in the geometric parameter space with the translation amount (dx, dy) as an axis is calculated, and the area information determination unit 1303 determines whether or not this intersection area is empty.
[0051]
As described above, the movable area in the (dx, dy) space is obtained from the following conditions.
[0052]
aj * dx + bj * dy- (cj + FMAX (j)) <= 0
Therefore, the intersection region is defined by selecting the smaller one of the FMAX (j) values for the boundary lines of the regions corresponding to the same ridge line.
[0053]
Whether or not the area is empty is the same as in the case of the movable area with respect to the single area, and is determined by the determination of the movable area and the presence / absence determination of the internal intersection shown in FIGS. .
[0054]
14 is a diagram showing an example of a composite area, FIG. 15 is a diagram showing a boundary line of the other area paired with FIG. 7, and FIG. 16 is a (dx, dy) space paired with FIG. The movable area in FIG.
[0055]
17 is obtained as a region where the composite region shown in FIG. 14 is present in the search image shown in FIG. 4 at the same time by the above-described flow of the composite region generation process shown in FIG. 12 and FIG. By proceeding with the determination of the composite area in this way, a composite area as shown in FIG. 18 can be obtained.
[0056]
The maximum value selection unit 106 in FIG. 1 calculates similarity for all composite regions detected by the composite region processing unit 104 described above from one image, and obtains the maximum value.
[0057]
Here, the similarity is calculated as follows.
[0058]
Similarity = 1− | Area of composite region−Area of search image | / Area of search image
The maximum value selection unit 106 sends the obtained maximum value to the ordering unit 107 as the similarity to the image. The ordering unit 107 receives the similarity to each image obtained from the maximum value selection unit 106 and selects k data in descending order of similarity.
[0059]
The data access unit 108 uses the similarity information of the sorted images stored in the internal memory of the ordering unit 107, reads a plurality of image files 109 from the one with the higher similarity, and displays it on the display 111.
[0060]
[Second Embodiment]
The second embodiment corresponds to the case where the size of the search image can vary. First, the principle will be described.
[0061]
In this second embodiment, when the enlargement rate of the search image is set to scale (reduction when less than 1) and the search image is enlarged around the origin, all points P (x, y) Move to (scale × x, scale × y). The equation of a straight line passing through the point P0P is
aj × x + bj × y + cj = 0
However, aj = y0−y,
bj = x-x0
cj = x0 * y-x * y0
Then, the equation of the straight line that passes through the point Q0Q which expanded this is
ak × x + bk × y + ck = 0
However, ak = scale × aj
bk = scale × bj
ck = scale × scale × cj
This is parallel to that before enlargement as a straight line expression. Therefore, the same applies to the point where the maximum value FMAX is given to the boundary row in the same region. That is,
FMAX (k) = ak * X + bk * Y = scale * (aj * X + bj * Y)
= Scale × FMAX (j)
An expression giving the inclusion condition
ak * dx + bk * dy- (ck + FMAX (k)) <= 0
Divided by scale
aj * dx + bj * dy- (scale * cj + FMAX) <= 0
This constitutes one side of the plane when considering a (dx, dy, scale) space.
[0062]
Similarly, the other ridge lines are also parallel, and the parameter regions that can exist are polyhedrons (in fact, one side is infinite) included in these plane groups.
[0063]
Therefore, by obtaining the intersection region of the polyhedron, it can be determined whether or not a plurality of regions can exist in the search image at the same time with the same magnification.
[0064]
FIG. 19 is a flowchart showing an embodiment of the composite area processing unit 104 that realizes the above principle.
[0065]
aj * dx + bj * dy- (scale * cj + FMAX) <= 0
Aj, bj, and cj are calculated as described above for all the ridge lines j of the search image, and FMAX is obtained by the operation of FIG.
[0066]
All intersections P = (dxi, dyi) are calculated as intersections of three non-parallel boundary surfaces, and it is determined whether or not this satisfies each conditional expression. The intersection of the three boundary surfaces can be obtained by solving the following simultaneous equations by a well-known simultaneous equation solving method when (i, j, k) is set to three different integers smaller than the number of edges of the search image. calculate.
[0067]
ai * dx + bi * dy- (scale * ci + FMAX (i)) = 0
aj * dx + bj * dy- (scale * cj + FMAX (j)) = 0
ak × dx + bk × dy− (scale × ck + FMAX (k)) = 0
Satisfaction of the conditional expression can be determined by the following expression.
[0068]
(Aj * dxj + bj * dyj-FMAX) / cj <= scale if cj> 0
(Aj * dxj + bj * dyj-FMAX) / cj <= scale if cj <0
If the scale can be increased indefinitely, an area of an arbitrary shape will be included. Therefore, practically, the range of the scale that can be changed is set to “1.2” or less, and the range is within this range. It is determined whether there is an internal intersection. That is,
(Aj * dx + bj * dy-FMAX) / cj <= 1.2 if cj> 0
(Aj * dx + bj * dy-FMAX) / cj <= 1.2 if cj <0
Thereby, it is possible to obtain an appropriate response even when the size varies.
[0069]
[Third embodiment]
The third embodiment of the present invention corresponds to a case where a search image is composed of a plurality of regions having different colors.
[0070]
Each pattern constituting the search image has the following information (vertex row of search image, vertical and horizontal width, area, color information of the image). In this case, the operation of the single area processing unit 102 shown in FIG. 3 is changed as described below.
[0071]
FIG. 20 is a diagram showing the configuration of the processing unit of the single area processing unit 102 in the third embodiment.
[0072]
The processing procedure corresponding to FIG. 3 is applied to the combination of individual pattern information and region information in the input search image, and the region description information and the search image can be moved in all cases where the conditions are satisfied. The area information is added to the single area memory 103. Thereafter, by configuring the processing unit of the composite area described above and the subsequent units, it is possible to perform a search using a search image using a plurality of patterns.
[0073]
In addition, when a part of an image is given as a search image, the search image is divided into regions, and each boundary line is approximated to a polygon to create a plurality of different color patterns. Pattern) can be generated. As a result, even when a part of an image is given as a search image using the above-described part, it is possible to search for image data including the inspection image pattern.
[0074]
The present invention may be applied to a system composed of a plurality of devices or an apparatus composed of a single device. The present invention can also be applied to a case where the object of the present invention is achieved by supplying a program for implementing the present invention to a system or apparatus.
[0075]
As described above, according to the present embodiment, by performing processing for a composite region in which a plurality of regions are combined together with processing for a single region, it is possible to correctly search for an object that is divided into a plurality of regions and extracted. There is.
[0076]
Further, there is an effect that the determination can be made efficiently by obtaining a movable area where each area and the search image have a specific phase relationship and determining the presence or absence of an intersection area of the movable area.
[0077]
【The invention's effect】
As described above, according to the present invention, there is an effect that the image data can be efficiently searched based on the target image extracted by dividing the image data into a plurality of regions.
[0078]
In addition, according to the present invention, there is an effect that an image can be searched based on image data divided in accordance with an area of image data recognized as one area.
[0079]
Further, according to the present invention, there is an effect that the time required for calculating the similarity of images can be shortened and images can be searched efficiently.
[0080]
[Brief description of the drawings]
FIG. 1 is a functional block diagram showing a functional configuration of an image search apparatus according to an embodiment of the present invention.
FIG. 2 is a diagram illustrating an example of an image search apparatus according to the present exemplary embodiment.
FIG. 3 is a functional block diagram showing a configuration of a single area processing unit in the first embodiment of the present invention.
FIG. 4 is a diagram schematically illustrating an example of a search image in the present embodiment.
FIG. 5 is a diagram schematically illustrating an example of a result of area division of image data to be searched.
FIG. 6 is a flowchart illustrating boundary line detection processing in a movable area calculation unit according to the embodiment.
7 is a schematic diagram for explaining a boundary line detected by the boundary line detection process of FIG. 6; FIG.
FIG. 8 is a diagram showing an example of a movable area of a single area in a geometric parameter space.
FIG. 9 is a flowchart illustrating determination processing of a movable region in the region information determination unit according to the embodiment.
10 is a flowchart showing a determination process for the presence or absence of an internal intersection in the determination process for the movable region in FIG. 9;
FIG. 11 is a diagram schematically showing a composite area inserted into a single area memory in the present embodiment.
FIG. 12 is a flowchart showing the flow of composite area generation and processing in a composite area processing unit.
FIG. 13 is a functional block diagram illustrating a configuration of a region information determination unit for each composite region in the composite region processing unit according to the first embodiment.
FIG. 14 is a diagram schematically illustrating an example of a composite region in the present embodiment.
FIG. 15 is a diagram showing a boundary line of a region detected by a boundary line detection process.
FIG. 16 is a diagram showing an example of a movable area of a single area in a geometric parameter space.
FIG. 17 is a diagram illustrating an example of a movable area of a composite area in a geometric parameter space.
FIG. 18 is a diagram schematically showing an example of a composite region obtained by this example.
FIG. 19 is a flowchart showing an internal intersection presence / absence determination process in the second embodiment.
FIG. 20 is a functional block diagram showing a functional configuration of a single area processing unit in the third embodiment;
[Explanation of symbols]
100 search images
101 Search image memory
102 Single area processing unit
103 single area memory
104 Compound area processing section
105 Multi-region memory
106 Maximum value selection section
107 Ordering unit
109 image files
110 Description file
111 display
301 Color information determination unit of single region processing unit
302 Size information determination unit of single region processing unit
303 Moveable area calculation unit of single area processing unit
304 Area information determination unit of single area processing unit
1301 Size information determination unit of composite area processing unit
1302 Intersection area calculation unit of compound area processing unit
1303 Area information determination unit of complex area processing unit

Claims (8)

An image search device for searching image data including an image similar to a specified image using image description information describing information for each region in the image as reference information,
A single area processing means for detecting a single area included in the search image from the individual areas in accordance with each of the individual areas in the image and a search image in which a search is specified;
A composite area processing means for obtaining a composite area combining the single areas included in the search image according to a combination with each of the single areas and the search image;
Similarity detection means for determining the similarity between the composite region and the search image;
Search means for searching for image data according to the similarity determined by the similarity detection means;
An image search apparatus comprising: The single area processing means determines the corresponding single area according to a difference in color information between each of the individual areas and the search image, and shape information of the search image. The image search device according to claim 1. The image search according to claim 1, wherein the single area processing unit determines the single area based on whether each of the individual areas and the search image have a predetermined positional relationship. apparatus. The composite area processing means determines a composite area based on an intersection area of geometric parameter areas indicating a range in which each of the individual areas and the search image are in a predetermined positional relationship. 2. The image search device according to 1. An image search method in an image search apparatus for searching image data including an image similar to a search image designated for search, from an image file ,
Storing a search image instructed to be searched in a search image memory;
The search in which each region in the image is stored in the search image memory with reference to a description file that describes information about the position, shape and color information of each region in the image stored in the image file A determination step of determining whether or not the image is included in the image ;
Storing the description information of the area determined to be included in the search image in the determination step, and the movable area of the search image with respect to the image in a single area memory;
Based on the description information and the movable area, if it is determined that each area stored in the single area memory satisfies the condition included in the search image at the same time , Seeking and storing in the composite area memory ;
A calculation step of calculating a similarity between the composite area stored in the composite area memory and the search image;
Searching image data including the search image according to the similarity calculated in the calculation step ;
An image search method characterized by comprising: In the area determined to be included in the search image in the determination step, the difference between the color information of the area and the color information of the search image is smaller than a predetermined value, and the shape of each of the individual areas is the search 6. The image search method according to claim 5, wherein the image search method is an area smaller than the shape of the image. The determined area to be included in the search image in the determination step further image according to claim 6, wherein the search image and the area is characterized the determined area der Rukoto with a predetermined positional relationship retrieval method. The composite area calculates a cross area of movable areas in a geometric parameter space with each of the areas and the parallel movement amount of the composite area as an axis, and the composite area is combined when the cross area exists in the search image . The image search method according to claim 5, wherein the image search method is determined to be an area .

Images