JP2005352990A - Partial image retrieval system, method and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a partial image retrieval system, method and program capable of reducing the information quantity of indexes to reduce a storage capacity necessary for holding the indexes while causing retrieval omission. <P>SOLUTION: In this partial image retrieval device for detecting a position containing a similar partial image from a group of accumulated images, indexes are assigned to accumulated partial area characteristics, respectively, accumulated partial area characteristics similar to an intended partial area characteristic are extracted as retrieval candidate characteristics by use of the indexes, the distance of each retrieval candidate characteristic from the intended partial area characteristic is calculated, and whether the similar partial image is present in the group of accumulated images or not is determined based on the distance. For a position of an accumulated image in which the partial image similar to the intended image is determined to be present in a collation result determination part, a collation window of the same size as an intended image is set to the accumulated characteristics to calculate the distance from the intended characteristic, and whether the partial image similar to the intended image is present in this position in the group of accumulated images or not is determined based on the characteristic distance. <P>COPYRIGHT: (C)2006,JPO&NCIPI

Description

  The present invention relates to a partial image search system and method suitable for identifying an image including a partial image similar to a given image, and a position where the partial image is included, from a plurality of stored images. .

With the spread of digital cameras, the increase in capacity and price of recording media, and the spread of large-capacity networks, an environment for easily acquiring and storing large-capacity digital images has been established.
For this reason, high-speed image search technology for quickly searching for a desired image from a large number of images is required. In particular, the partial image search is a part of the image in which the image to be searched is stored. In addition, higher speed is required.
This partial image retrieval technique is used for an image detection technique for retrieving an image including a target image as a part thereof from a huge image database at high speed.

For example, a homepage that includes an image in which a face of a specific person is captured from images stored in an image database and presented to the user, or an image in which a specific object or building is captured It is used for search technology that searches for.
Conventionally, with regard to the partial image search method, a high-speed partial image detection method for accurately searching for a location of an image similar to an image to be searched as a sample from an image of a larger size without omission is an object detection device (for example, a patent This method is known as a high-speed template matching method (see Non-Patent Document 1, for example).

However, in the above-described method, since all the partial regions of all the images are collated and searched regardless of the similarity to the target image, a lot of useless collating processes occur, and the search is performed at high speed. There is a drawback that it can not be made.
Therefore, as a method for speeding up the search for partial images, it is conceivable to index each image in advance in order to easily search for images.
However, when all the partial images that can be located are extracted from each image, the number of the images becomes enormous, and if all these images are indexed, the storage capacity necessary for index retention becomes enormous. There is a point.

On the other hand, a partial image detection method that reduces the number of partial images to be indexed by clustering partial images as a method for solving the above-described problem has been known (for example, see Non-Patent Document 2). ing.
JP 09-330404 A "High-speed template matching method using residual information between partial regions of target image", Takahito Kawanishi, et al, Information Science and Technology Forum, Proceedings, LI-008, September 2003 "WALRUS: A similarity retrieval algorithm for image databases", A. Natsev, et al, IEEE Transactions on Knowledge and Data Engineering, Vol. 16, No-3, March 2004

However, in the partial image detection method for clustering the partial images shown in Non-Patent Document 2 described above, the number of partial images can be reduced compared to Patent Document 1 and Non-Patent Document 1, but the capacity reduction can be achieved. From the viewpoint, the number of partial images is not sufficiently reduced, and the problem that the storage capacity necessary for holding the index becomes enormous has not been solved.
Further, the method of Non-Patent Document 2 has a drawback that the calculation time required for clustering the partial images is very long, so that the time required for assigning the index becomes enormous.
The present invention has been made in view of such circumstances, and a partial image search system, method, and program for reducing the amount of information in an index, reducing the storage capacity necessary for holding the index, and causing no search omission. The purpose is to provide.

  The partial image search method of the present invention is a partial image search method for detecting an image including a partial image similar to a target image to be searched from a stored image group registered in a database, and a position of the partial image in the image. The accumulated feature extraction process for extracting the accumulated feature that is a feature of each image of the accumulated image group, and the feature in the attention window at each position are stored in the accumulation portion while shifting the attention window of a predetermined size at a predetermined interval. A storage partial region feature extraction process for extracting as a region feature; an indexing step for assigning an index for labeling each of the storage partial region features; a target feature extraction step for extracting a target feature that is a feature of the target image; In the target feature, a target window of a predetermined size is set, and the target partial region feature that extracts the feature in the target window at each position as the target partial region feature while shifting the target window. An extraction process, an index search process for extracting a stored partial area feature similar to the set of target partial area features as a search candidate feature using the index, and a distance between the search candidate feature and the target partial area feature. A feature matching process to be calculated; a matching result determination process for determining whether or not there is a possibility that a partial image similar to the target image exists at the location in the accumulated image group based on the distance; and the matching result In the determination process, a matching window having the same size as the target image is set for the stored feature at a location in the stored image where it is determined that there is a possibility that a partial image similar to the target image exists. A feature re-matching process for calculating a feature distance with the target feature, and a matching result re-check for determining whether or not a partial image similar to the target image exists at the location in the stored image group based on the feature distance. Overjudgment A search result candidate selection process in which a location in the stored image determined that a partial image similar to the target image exists in the matching result re-determination process is a search result candidate; and a set of the target partial region features Next target partial area feature selection process for designating a target partial area feature to be matched next, and next search candidate feature for designating a storage partial area to be matched next from the set of the accumulated partial area features And a selection process.

  As described above, according to the present invention, in each image of the accumulated image group, the partial image is extracted by thinning out the partial region at predetermined intervals narrower than the partial region, and an index is assigned to each pixel. Compared to a known method such as Non-Patent Document 2 that extracts a partial image and assigns an index to each image of the accumulated image group without shifting, without increasing the time required for indexing, Since the storage capacity required to maintain the index is reduced and all partial areas are extracted from the target image without any interval, the search failure of the stored image occurs. The effect that it can be guaranteed not to be obtained is obtained.

<First embodiment>
Hereinafter, a partial image search apparatus according to a first embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing an example of the configuration of the partial image search apparatus of the first embodiment.
In this figure, the partial image search apparatus according to the present embodiment includes an accumulation feature extraction unit 1, an accumulation partial region feature extraction unit 2, an indexing unit 3, a target feature extraction unit 4, and a target partial region feature extraction unit 5. The index search unit 6, the feature matching unit 7, the matching result determining unit 8, the feature re-matching unit 9, the matching result redetermining unit 10, the search result candidate selecting unit ll, and the next target partial region feature selecting unit. 12 and a next search candidate feature selection unit 13, which are a target image, that is, an image to be searched that is referred to, and a stored image group that is stored in a database or the like, that is, a searched image to be searched. A set of accumulated images is input, and a location in the accumulated image containing a partial image similar to the target image is output.

The accumulated feature extraction unit 1 extracts accumulated features representing an image from each image of the accumulated image group in the entire image or a block unit having a predetermined number of pixels.
The accumulation partial region feature extraction unit 2 sets a predetermined window of a predetermined size with respect to the feature output from the accumulation feature extraction unit 1 at a predetermined interval (a predetermined interval). And the accumulated partial region feature which is a feature in each window of interest is extracted.
The index assigning unit 3 vector-quantizes each accumulated partial region feature output from the accumulated partial region feature extracting unit 2, and assigns a vector quantization codeword as an index for a sign.

The target feature extraction unit 4 extracts a target feature that represents an image from the target image.
The target partial region feature extraction unit 5 sets a target window of a predetermined size for the target feature extracted by the target feature extraction unit 4 and shifts the target window by one pixel. The target partial region feature that is the feature in the window of interest is extracted.
The index search unit 6 extracts similar accumulated partial region features from the set of target partial region features extracted by the target partial region feature extraction unit 5 using the index output from the index assigning unit 3.

The feature matching unit 7 calculates the distance between the accumulated partial region feature extracted by the index search unit 6 and the target partial region feature extracted by the target partial region feature extraction unit 5.
As this distance, Euclidean distance, Manhattan distance, inner product, normalized cross-correlation, and the like can be used.
Based on the distance calculated by the feature matching unit 7, the matching result determination unit 8 determines whether or not there is a possibility that a partial image similar to the target image exists in the location in the accumulated image group.

When the matching result determination unit 8 determines that there is a possibility that a partial image similar to the target image exists, the feature re-collation unit 9 extracts the accumulated feature extraction unit 1 at the determined location in the accumulated image. A matching window having the same size as the target image is set for the stored features, and a feature distance from the target feature output from the target feature extraction unit 4 is calculated. Various definitions can be used for this characteristic distance as well as the above distance.
The collation result re-determination unit 10 determines whether or not a partial image similar to the target image exists at the location in the accumulated image group based on the distance used in the collation result determination unit 8.

The search result candidate selection unit 11 sets a location in the accumulated image, which is determined by the matching result re-determination unit 10 as having a partial image similar to the target image, as a search result candidate for performing detailed similarity examination.
The next target partial region feature selection unit 12 designates a target partial region feature to be collated next from the set of target partial region features extracted by the target partial region feature extraction unit 5.
The next search candidate feature selecting unit 13 designates a storage partial region to be collated next from the set of storage partial region features output as a search result from the index search unit 6.

Next, the operation of the partial image search apparatus described above will be described with reference to FIGS. FIG. 2 is a flowchart specifically showing an operation example of the partial image search apparatus according to the first embodiment of FIG.
First, the accumulated feature extraction unit 1 reads an accumulated image group from a database (not shown) (step S1).
Next, the accumulated feature extraction unit 1 extracts an accumulated feature that is a feature of the image from each image of the inputted accumulated image group (step S2).

  For example, if the RGB pixel values are used as they are and the size of the accumulated image is “384 × 256” pixels, the accumulated features (from multidimensional vectors) composed of “384 × 256 × 3 = 294912” numerical values. For each accumulated image, that is, one per image. Here, “3” multiplied by “384 × 256” is the number of RGB colors.

As another embodiment, for example, a small area (block) having a size of about 8 × 8 pixels is used from an image, and the small areas are sequentially shifted by 1 pixel in the vertical and horizontal directions. It can be obtained by extracting the accumulated features within the range and calculating the average value of each RGB of each small region.
That is, assuming that the size of the accumulated image is “384 × 256” pixels, a feature composed of “(384-8 + 1) × (256-8 + 1) × 3 = 281619” numbers is obtained for each accumulated image. In other words, one image is obtained.

Furthermore, as another embodiment, for example, a small area having a size of about 8 × 8 pixels is shifted from the accumulated image by 1 pixel in the vertical and horizontal directions, and each of RGB of pixels included in each small area is shifted. Two-dimensional DCT (digital cosine transform) can be performed, and the coefficient vector can be used as an accumulation feature.
That is, assuming that the size of the accumulated image is “384 × 256” pixels, an 8 × 8 × 3 = 192-dimensional feature vector is “(384-8 + 1) × (256-8 + 1) = 93873 ”pieces are calculated.
In this case, the accumulated feature extraction unit 1 outputs a set of accumulated features that are the features extracted from the accumulated image.

Next, the storage partial region feature extraction unit 2 reads a set of storage features from the storage feature extraction unit 1 and sets a first target window of a predetermined size at a certain location of the storage image. To do.
For example, the first window of interest has a size of 64 × 64 pixels, and is arranged in the upper left corner of the stored image as an initial state. .
Then, the storage partial region feature extraction unit 2 extracts a storage partial region feature that is a feature of the image of the partial region included in the set first attention window.
For example, for each of the accumulated partial area features, the RGB pixel value is used as it is, and one “64 × 64 × 3 = 1288 dimensional feature vector is obtained.

In another embodiment, for example, an average of RGB surface element values in a small region is calculated as the accumulated partial region feature, and the obtained average value is used to overlap the first attention window. The above-mentioned small areas are arranged in a state where there is no gap, and features corresponding to the small areas are extracted, and a feature of “(64/8) × (64/8) × 3 = 192” dimension is extracted. It is also possible to obtain one vector.
Further, in another embodiment, for example, as the accumulated partial region feature, the two-dimensional DCT coefficient vector in the small region is used, and the small region is not overlapped in the first attention window and there is no gap. Areas are arranged, and features corresponding to the respective small areas are extracted, and “(64/8) × (64/8) = 64” feature vectors of “8 × 8 × 3 = 192” dimensions are obtained. Alternatively, all the dimension feature vectors can be combined into one to obtain one “192 × 64 = 1288” dimension feature vector.

A plurality of first attention windows are set at predetermined intervals in each accumulated image, and the accumulated partial area features are extracted from the partial areas in each attention window (step S3).
That is, the first window of interest (horizontal ωx pixel × vertical ωy pixel) is the predetermined interval (interval between the same sides of each window of interest) in each accumulated image in the vertical and horizontal directions. , The interval between the upper sides is sequentially shifted by mx pixels in the horizontal direction and my pixels in the vertical direction), and each of the overlapping first windows is shown in FIG. In this way, they overlap except for a predetermined interval.

For example, if the first window of interest is arranged at intervals of 16 pixels vertically and horizontally, and the size of the accumulated image is “384 × 256” pixels, the first window of interest is “64 × 64” pixels. For each image, “21 × 13 = 273” first attention windows are set.
Further, the arrangement interval of the first window of interest is hereinafter referred to as a margin. For example, if the margin is set to be equal to or smaller than the size of the first target window, the first target window is always arranged so as not to cause a search omission.
Then, the storage partial region feature extraction unit 2 outputs a set of storage partial region features obtained for each of the first windows of interest.

Next, the index assigning unit 3 reads the set of accumulated partial region features extracted by the accumulated partial region feature extracting unit 2, and assigns an index for facilitating the search to each read partial region feature. .
For example, as in the embodiment described in “Signal Detection Method, Signal Detection Device, Recording Medium, and Program” (Japanese Patent Application Laid-Open No. 2002-236696), a feature of each partial region image is created using a vector quantization codebook created in advance. The vector is vector-quantized, and the codeword of vector quantization is used as an index.

That is, the above method will be briefly described. First, the index assigning unit 3 reads a set of accumulated partial region features.
Next, the index assigning unit 3 performs clustering of each accumulated partial region feature. This clustering is performed by, for example, considering each accumulated partial region feature as a vector and encoding the vector using vector quantization.
For example, if the number of code words for vector quantization is 1024, the partial region features are classified into any of 1024 clusters.

Here, the vector quantization uses, for example, a square error as a distance function, assigns a partial region feature to a cluster whose distance from the cluster centroid is smaller than the distance from any other cluster centroid, and codes corresponding to the clusters. This is done by outputting a word.
This cluster is created in advance using, for example, learning features that are prepared in advance for cluster creation and have the same number of dimensions as the partial region features.
That is, a cluster is such that the sum of the distance between the learning feature belonging to it and the cluster centroid is minimized, and the distance from the cluster centroid of the cluster to which the cluster belongs is any other cluster. It is comprised so that it may become smaller than the distance with a gravity center.
Then, the index assigning unit 3 outputs a vector quantization code word corresponding to each partial region feature as an index (step S4).

As another embodiment, for example, R ^* -Tree (“The R ^* -Tree: an efficient and rbust access method for points and rectangles”, (N. Beckman et al., Proc. Of ACM SIGMOD Conference, pp .322-331, 1990)) and SR-Tree ("SR-Tree: Proposal of index structure for nearest neighbor search for high-dimensional data" (Natao Katayama et al., IEICE Transactions DI, Vol. J80-DI, No.8, pp.703-717 (August 1997)) and other methods that are collectively referred to as multidimensional indexes It is also possible to create an index and assign an index by associating the lowest layer area with a feature.
The region including the feature is, for example, a super cuboid corresponding to a rectangular parallelepiped in three-dimensional space if R ^* -Tree, and a super sphere corresponding to a sphere in three-dimensional space if SR-Tree. It becomes an overlap.
Then, the index assigning unit 3 outputs an index for each partial region feature obtained as described above.

Next, the target feature extraction unit 4 reads the target image (step S5). If the size of the target image is set to be equal to or larger than the sum (total) of the margin and the size of the first window of interest, for example, Since the range in which the feature of the stored image is completely extracted is included, no omission of search occurs.
Then, the target feature extraction unit 4 performs feature extraction from the target image by the same method as the accumulated feature extraction unit 1, and outputs a set of target features that are the features extracted from the target image (step S6). .

Next, the target partial region feature extraction unit 5 reads the set of target features extracted by the target feature extraction unit 4, and sets the target window used by the storage partial region feature extraction unit 2 in one location of the target image. A target window having the same size (vertical × horizontal), that is, a first target window is used, and a target partial region feature that is a feature of an image included in the first target window is extracted.
The target partial region feature extraction unit 5 extracts image features using the same extraction method as the storage partial region feature extraction unit 2 and outputs the extraction result as a set of target partial region features (step S7).

Here, unlike the case of the feature extraction in the storage partial region feature extraction unit 2, the target partial region feature extraction unit 5 does not overlap the first target window in the target image and has no gap. The target partial region features are extracted at each position while setting them and shifting them one pixel at a time.
At this time, for example, if the first extraction window is shifted at least by the size of the margin, no search omission occurs.
For example, when the size of the target image is “80 × 80” pixels, the size of the first target window is “64 × 64” pixels, and the margin is “16 × 16” pixels, the target image may overlap. Since the number of first target windows that can be set in a state without any gap is one, “16 × 16 = 256” first target windows are set in the entire target image.
Then, the target partial region feature extraction unit 5 outputs a set of target partial region features extracted by the above-described processing.

Next, the index search unit 6 reads the index output from the index assigning unit 3 and the set of target partial region features output from the target partial region feature extraction unit 5.
When the target representative feature extraction unit 15 performs the process of extracting the target representative feature from the target partial region feature, the index search unit 6 uses the target representative feature extraction unit 15 instead of the set of target partial region features. Read a set of target representative features to be output.
The target representative feature extraction unit 15 will be described in detail in another embodiment to be described later.
In addition, the index search unit 6 may further read the target inter-region distance output from the target inter-region distance calculation unit 16 as necessary. The target partial region distance calculation unit 16 will be described in detail in another embodiment described later.
Further, when the search threshold update unit 14 performs a process of updating the search threshold and the selection threshold, the index search unit 6 reads the updated selection threshold. The search threshold update unit 14 will be described in detail in another embodiment described later.

Then, the index search unit 6 uses the read index to extract an accumulated partial area feature similar to the target partial area feature set as a search candidate feature (step S8).
At this time, the index search unit 6 selects, for any target partial region feature in the set of target partial region features, a distance dt from the partial region feature, for example, a square error, which is a predetermined threshold value. Accumulated partial region features that are below θ ₂ (below) are extracted by index.
The selection threshold θ ₂ is a value that is automatically determined from parameters described later, and a method for determining the selection threshold will be described later.

As another example, instead of a set of target sub-region features, a representative feature is extracted from the target sub-region feature as a target representative feature, and this target representative feature set is used to distance dt between objects representative feature is the storage part region feature below a selected threshold theta ₂ of, it is also possible to adopt a configuration that extracts from the index (described later).
Further, as another embodiment, the distance between the target partial areas is used by using the distance between the target partial areas, and the distance dt between each target representative area is reduced by the maximum value of the distance between the target partial areas corresponding to the target representative feature. It is also possible to adopt a configuration in which accumulated partial region features whose values are below the selection threshold θ ₂ are extracted from the index (described later).

Here, an extraction method for extracting from the index an accumulated partial region feature whose distance from a certain target partial region feature or target representative feature is below the selection threshold will be described.
This extraction method differs depending on the type of index used in the index assigning unit 3 and can be implemented in the same manner as the method described in the patent specification or the method described in the non-patent document corresponding to the index.
For example, when an index is assigned using the method described in “Signal Detection Method, Signal Detection Device, Recording Medium, and Program” (Japanese Patent Laid-Open No. 2002-236696), a desired accumulated partial region feature is extracted as follows. be able to.

First, the index search unit 6 reads and purpose partial region feature or objects representative feature, indexes created by the indexing portion 3 (the cluster), and the selection threshold theta _2.
In the following, for the sake of simplicity, the description will be made assuming that the target partial region feature is read in the configuration adopted in this embodiment, not the target representative feature.
Then, the index search unit 6 calculates the distance between the read target partial region feature and each cluster centroid.
Subsequently, the index search unit 6, based on the calculated distance dt, selects a potential cluster including storage part region feature where the distance between the objective portion region feature is below a selected threshold theta _2.

The principle of selecting the above-described cluster will be described below.
FIG. 4 is a plane on which three points Q, C1, and C2 are placed, and shows a state in which a feature space in which feature vectors are arranged is cut out.
Here, the point Q is the target partial region feature, C1 is the cluster centroid of the cluster to which the partial region feature Q belongs, C2 is the cluster centroid of the other cluster, and the distances dQ1, dQ2, and d12 are the points. The distance between Q and point C1, the distance between point Q and point C2, and the distance between point C1 and point C2 are shown.

Here, if it is necessary to extract an accumulated partial region feature whose distance dt from the point Q that is the coordinates of the target partial region feature is within the selection threshold θ ₂ , a hypersphere having a radius θ ₂ centered on the point Q storage part region feature in the interior of true thereto (circle of radius d _theta in FIG. 4).
Then, when the radius of the hypersphere centered at the point Q becomes larger than _dθ in FIG. 4, a desired storage partial region feature is included in the storage partial region features belonging to the cluster corresponding to the point C2. There is a possibility.
Therefore, when the selection threshold θ ₂ becomes larger than d _θ in FIG. 4, the cluster is selected. _dθ is obtained using the following equations (1) and (2).

  From the above equation (1),

Is required.
Finally, when the following expression (3) holds, the index search unit 6 selects a cluster corresponding to the point C2, and extracts all accumulated partial region features belonging to the cluster.
Then, this procedure is performed for all clusters except the cluster to which the target partial region feature Q belongs, and the extracted accumulated partial region features are output as search candidate features.

Next, the feature matching unit 7 reads the set of target partial region features output from the target partial region feature extraction unit 5 and the set of search candidate features output from the index search unit 6.
The feature matching unit 7 reads the target partial region feature when the next target partial region feature selecting unit 12 specifies the target partial region feature to be verified next. The next target partial region feature selection unit 12 will be described later in this embodiment.

Further, when a search candidate feature to be matched next is designated from the next search candidate feature selection unit 13, the feature matching unit 7 reads the search candidate feature. This next search candidate feature selection unit 13 will be described later in this embodiment.
Next, the feature matching unit 7 performs matching between the target partial region feature read from the next target partial region feature selecting unit 12 and the set of search candidate features read from the next search candidate feature selecting unit 13.
At this time, the feature matching unit 7 calculates a distance between the target partial region feature and the search candidate feature, for example, a square error.

Here, if there is no designated target partial region feature, the feature matching unit 7 designates, for example, a target partial region feature corresponding to the position of the upper left corner of the target image as the initial target partial region feature, and the search candidate A feature and matching process is performed (step S9).
Further, if there is no designated search candidate feature, the feature matching unit 7 designates, for example, an arbitrarily selected search candidate feature as an initial search candidate feature, and collates it with the target partial region feature.
Then, the feature matching unit 7 calculates the distance value d (ω) between the matched search candidate feature and the target partial region feature as a result of the matching calculation (calculation of the distance between the search candidate feature and the target partial region feature). Output.

Next, the matching result determination unit 12 reads the distance value d (ω) output from the feature matching unit 7 and determines whether or not there is a possibility that a partial image similar to the target image exists at the current matching location. That is, the distance value d (ω) is compared with the selection threshold value θ ₂ .
The verification result determining section 8, when the distance value d (omega) is equal to or less than the selected threshold theta _2, it determines that there is a possibility that the partial image similar to the target image exists in the current matching point, outputting the stored partial region feature as the matching portion, whereas, when the distance value d (omega) exceeds the selection threshold theta _2, if there is no possibility of partial image similar to the target image is in the current verification point Determination is made and the process proceeds to step S14 (step S10).

Next, the feature re-collation unit 9 includes a set of accumulated features output from the accumulated feature extraction unit 1, a set of target features output from the target feature extraction unit 4, and a collation output from the matching result determination unit 8. Read the location.
Then, the feature re-collation unit 9 sets a second attention window, which is the attention window having the same size (vertical × horizontal) as the target image, in the read accumulated feature.

Here, the feature re-collation unit 9 determines the location of the second window of interest corresponding to the collation location output from the collation result determination unit 8.
That is, as shown in the conceptual diagram of FIG. 5, the feature re-collation unit 9 has the same relative position of the accumulation partial area with respect to the arrangement position of the second target window as the relative position of the target partial area in the target image. In other words, the second window of interest is arranged at a position where the accumulation partial area and the target partial area at the collation location overlap.

Next, the feature re-collation unit 9 collates the accumulated feature and the target feature in the second attention window.
Here, the feature re-collation unit 9 calculates, for example, a distance value d between the accumulated feature in the second attention window and the target feature, for example, a square error.
Then, the feature re-collation unit 9 outputs a distance value d that is a collation calculation result (step S11).

Next, the collation result re-determination unit 14 reads the distance value d output from the feature re-collation unit 13, and determines whether or not there is a possibility that a partial image similar to the target image exists in the current collation location. That is, the distance value d is compared with the search threshold value θ which is the predetermined threshold value.
At this time, when the distance value d is equal to or less than the search threshold θ, the matching result re-determination unit 10 determines that a partial image similar to the target image exists at the current matching location, and detects the matching location in the accumulated image as a detected location. The position and determination result (information indicating the presence) of this detection location is output, and when the distance value d exceeds the search threshold θ, a partial image similar to the target image exists at the current verification location It determines with not, and advances a process to step S14 (step S12).

The selection threshold θ ₂ is determined from the search threshold θ, and if the selection threshold θ ₂ is set as shown in the following equation (4), no search omission occurs.

  In the above equation (4), N is the minimum of the first matching window included in the second matching window placed at an arbitrary position when it is assumed that the first window of interest is placed in the stored image without overlapping or gaps. It is a number and is given by the following equations (5) to (7).

Here, Nx in the equation (6) is the number of first collation windows in the horizontal direction, and Ny in the equation (7) is the number of first collation windows in the vertical direction.
However, (qx, qy) is the size of the target image, that is, the size of the second target window, and (ωx, ωy) is the size of the first target window.
Next, the search result candidate selection unit 11 reads the detection location output from the verification result re-determination unit 10 and the distance value d output from the feature re-verification unit 9.

Next, the search result candidate selection unit 11 registers the detection location and the distance value d at the detection location as search result candidates.
For example, the search result candidate selection unit 11 registers all the combinations of the read detection location and the distance value d regardless of the distance value.
In addition, as another embodiment, the search result candidate selection unit 11 may be configured to register the search result candidates so that the number of search result candidates is equal to or less than a predetermined number.

That is, the search result candidate selection unit 11 unconditionally registers the search result candidates up to the quantity, and sets a combination of the detected location and the distance value. If the quantity exceeds the quantity, for example, the search result candidate already read The maximum distance value dmax in the inside is compared with the distance value d read thereafter, and when the read distance value d falls below the maximum distance value dmax, the candidate corresponding to the maximum distance value dmax is deleted and newly read detection A location and a set of distance values d are registered as candidates.
As described above, the search result candidate selection unit 11 outputs the search result candidate from the accumulated image by the method described above (step S13).

Next, the next target partial region feature selection unit 12 reads the set of target partial region features output from the target partial region feature extraction unit 5 and, if necessary, the distance value d ( ω), the selection threshold θ ₂ output from the search threshold update unit 14, and the peripheral distance lower limit value output from the peripheral distance lower limit calculation unit 17 are read.
The search threshold update unit 14 and the peripheral distance lower limit calculation unit 17 will be described later in another embodiment.
Then, the next target partial region feature selection unit 12 detects the presence or absence of the target partial region feature to be collated next. When it is detected that there is no collation location, the process proceeds to step S15. If detected, the target partial region feature to be checked next is selected, and the process proceeds to step S9 (step S14).

At this time, for example, the next target partial region feature selection unit 12 determines whether or not there is a target partial region feature that has not been collated with respect to the storage partial region feature currently focused on, and ends. If there is a target partial area feature that is not yet selected, one arbitrary target partial area feature is selected from these, and when all target partial area features have been collated, attention is paid to the next storage partial area feature. In order to change, the process proceeds to step S15.
As another embodiment, the following objective partial region feature selection unit 12, if read the peripheral distance limit value, the distance limit value, omit collation purposes partial region feature above selected threshold theta ₂ read and, it is also possible to adopt a configuration in which the distance limit value is any one to choose for the selection threshold theta ₂ following purposes partial region feature.
Then, if there is an unfinished target partial region feature, the next target partial region feature selection unit 12 outputs the target partial region feature whose collation location is selected and the position in the target image corresponding to the target partial feature. To do.

Next, the next search candidate feature selection unit 13 reads a set of search candidate features output from the index search unit 6, selects a search candidate feature to be matched next, that is, whether there is a search candidate feature to be matched next. When detection is performed and it is detected that there is no search candidate feature to be matched, the process is terminated. On the other hand, when it is detected that there is a search candidate feature to be matched, a search candidate feature to be matched next is selected and the process is performed. The process proceeds to S9 (step S15).
At this time, for example, when detecting that there is a search candidate feature to be collated, the next search candidate feature selection unit 13 selects one arbitrarily from the search candidate features that have not been collated.
As a result, the next search candidate feature selection unit 13 outputs the selected search candidate feature to the feature matching unit 7.

Second Embodiment (Corresponding to Claim 3)
FIG. 6 is a block diagram showing a configuration example of a partial image retrieval apparatus according to a second embodiment to which the method according to claim 3 is applied.
In addition to the partial image search apparatus according to the first embodiment to which the method of claim 1 is applied, the partial image search apparatus according to the second embodiment further includes a search threshold update unit 14 as shown in FIG. A partial image that contains an image that is similar to the target image, with an input of an arbitrary target image, that is, a sample image to be searched, and a set of stored images, that is, a set of stored images that are a plurality of searched images. The location in the stored image is output.

The search threshold update unit 14 detects candidates within a predetermined distance from the search result candidates output from the search result candidate selection unit 11 based on the distance value output from the feature re-collation unit 9. The search threshold θ is newly set based on the detected candidate distance value.
Next, the partial image detection processing from the accumulated feature extraction unit 1 to the search threshold update unit 14 will be described in detail with reference to FIG. Here, FIG. 7 is a flowchart showing an operation example of the partial image search apparatus shown in FIG.

In the flowchart of FIG. 7, the “search threshold update” process of step S16 is added to the flowchart of the first embodiment shown in FIG. 2 after the “search result candidate selection” process of step S13. Since the other processes are the same, only the process of step S16 will be described.
The search threshold update unit 14 reads the search result candidate output from the search result candidate selection unit 11, selects the search result candidate having the maximum distance value from the read search result candidates, and selects the selected search result candidate. It is determined whether or not the distance value is equal to or less than the search threshold value θ. When it is detected that the distance value is equal to or less than the search result threshold value θ, the search threshold value θ is updated to the same value as the distance value.
Then, (4) the selection threshold theta ₂ from the search threshold theta calculated based on equation outputs with updated search threshold theta (step S14).
Thereafter, the updated search threshold θ and selection threshold θ ₂ are used for determination of the collation result.

<Third embodiment> (corresponding to claim 5)
FIG. 8 is a block diagram showing a configuration example of a partial image search apparatus according to a third embodiment to which the method according to claim 5 is applied.
The partial image search device according to the third embodiment is obtained by adding a target representative feature extraction unit 15 to the partial image search devices of the first and second embodiments, and serves as a target image, that is, a sample. Input an image to be searched and a stored image group, that is, a set of stored images that are to be searched, and search for a location in the stored image of a partial image containing an image similar to the target image, Output search results.
The target representative feature extraction unit 15 extracts a target representative feature as a feature common to each target partial region feature from the set of target partial region features output from the target partial region feature extraction unit 5.

Next, the partial image detection processing in the above-described accumulated feature extraction unit 1 to target representative feature extraction unit 15 will be specifically described with reference to FIG. Here, FIG. 9 is a flowchart showing an operation example of the partial image search apparatus shown in FIG.
The flowchart of FIG. 9 is different from the flowchart of the second embodiment shown in FIG. 7 in that the process of “extraction of target representative features” in step S17 is followed by the process of “extraction of target partial area” in step S7. Since only the process is added and the other processes are the same, only the process of step S17 will be described.

The target representative feature extraction unit 15 reads a set of target partial region features output from the target partial region feature extraction unit 5 and performs a clustering process on the plurality of target partial region features.
At this time, for example, the target representative feature extraction unit 15 prepares only one cluster (extracts a common feature vector), and assigns all target partial region features in the set to the cluster (one target representative). Will output features).

Further, as another example, the target representative feature extraction unit 15 extracts the positions in the target image corresponding to the target partial region features for all the read target partial region features, and divides the existence range into a grid pattern. It is also possible to adopt a configuration in which each lattice is associated with a cluster.
In addition, as another example, the target representative feature extraction unit 15 may prevent the maximum value of the distance between target partial region features in the same cluster from exceeding a classification threshold that is a predetermined threshold. Can also be merged.

At this time, the target representative feature extraction unit 15 uses, as the initial state of the cluster, for example, one in which target partial areas are assigned to clusters prepared for the number of read target partial area features.
Next, the target representative feature extraction unit 15 extracts a representative feature that is a partial region feature representing the cluster from each cluster. The target representative feature extraction unit 15 uses, for example, one arbitrarily selected target partial region feature in the cluster as a representative feature.

In addition, as another example, the target representative feature extraction unit 15 can use a target partial region feature corresponding to a specific position as a representative feature.
This specific position is, for example, the position of the upper left corner of the target image.
Further, as another example, the target representative feature extraction unit 15 may be configured such that the target partial feature at the center of gravity of the target partial region feature in the cluster is a representative feature.
Then, the target representative feature extraction unit 15 outputs the representative feature or the set obtained by each embodiment in the third embodiment described above.
In the next step S8, the index search unit 6 performs a matching process with the index of the stored image using the target representative feature.

<Fourth embodiment> (corresponding to claim 7)
FIG. 10 is a block diagram showing a configuration example of a partial image search apparatus according to a fourth embodiment to which the method according to claim 6 is applied.
The partial image search device according to the fourth embodiment is different from the partial image search devices according to the first, second, and third embodiments in the target partial region distance calculation unit 16 and the peripheral distance lower limit value calculation unit 17. The target image, that is, the image to be searched as a sample, and the stored image group, that is, the set of stored images that are the searched images, are input, and an image similar to the target image is included. The part of the stored partial image is searched for in the stored image, and the search result is output.

The target inter-region distance calculation unit 16 calculates the distance between the partial region features for the target partial region feature output from the target partial region feature extraction unit 5.
If the target representative feature extraction unit 15 is provided, the distance between the target partial region feature and the corresponding target representative feature is calculated.
The peripheral distance lower limit calculation unit 17 uses the distance value output from the feature matching unit 7 and the distance value output from the target inter-region distance calculation unit 16 to perform a search performed by the feature matching unit 7. A distance between the candidate feature and each target partial region feature output from the target partial region feature extraction unit 5 is calculated, and a lower limit value is extracted from the calculation result.

Next, the partial image detection processing from the accumulated feature extraction unit 1 to the peripheral distance lower limit calculation unit 17 will be described in detail with reference to FIG. Here, FIG. 11 is a flowchart showing an operation example of the partial image search apparatus shown in FIG.
The flowchart of FIG. 11 is different from the flowchart of the third embodiment shown in FIG. 9 in the “calculation of distance between target partial areas” in step S18 after the process of “extraction of target representative features” in step S17. A process is added, and after the process of “update search threshold” in step S16, only the process of “calculation of lower limit of peripheral distance” in step S19 is added, and the other processes are the same. Therefore, only the processing in steps S18 and S19 will be described.

The target partial region distance calculation unit 16 reads a set of target partial region features output from the target partial region feature extraction unit 5 and calculates a distance between the target partial region features.
Then, the target inter-region distance calculation unit 16 performs distance calculation for all sets of target sub-region features, for example. Here, for example, a square error is used as the distance scale.
As another example, when the target representative feature extracting unit 15 is provided, a set of target representative features output from the target representative feature extracting unit 15 is further read, and each target partial region feature, It is also possible to calculate the distance to the corresponding target subregion feature.
As described above, the target partial region distance calculation unit 16 outputs the distance between the target partial region features or the distance between the target partial region feature and the target representative feature.
Using the distances described above, an index search is performed in step S8.

Next, the peripheral distance lower limit value calculation unit 17 reads the distance value output from the feature matching unit 7 and the inter-subregion distance output from the target inter-region distance calculation unit 16.
Then, the peripheral distance lower limit value calculation unit 17 calculates a distance lower limit value between the accumulated partial area feature and each target partial area feature from the read distance value and the partial area query distance.
At this time, the peripheral distance lower limit value calculation unit 17 sets the lower limit values d (f _D ^(W) and f _Q2 ^{(W )} of the distance between the accumulated partial region feature f _D ^(W) and the target partial region feature f _Q2 ^{(W). )} ) Is obtained by the following equation (8) based on the triangular inequality. ,

In the above equation (8), d (f _D ^(W) , f _Q1 ^(W) ) is a distance value read from the feature matching unit 7, and d (f _Q1 ^(W) , f _Q2 ^(W) ) is read. This is the distance between the target partial areas.
As described above, the peripheral distance lower limit value calculation unit 17 outputs the peripheral distance lower limit value that is the lower limit value of the distance between the accumulated partial region feature and the target partial region feature.
The omission in the next step S14, the next objective partial region feature selection unit 12, and inputs the near distance limit value, the distance limit value, the collation of an object part region feature above a selected threshold theta ₂ read and, this distance limit value is any one to choose for the selection threshold theta ₂ following purposes partial region feature.

<Experiment>
An example of the operation of the apparatus to which the present invention is applied is shown below.
Here, 1000 images with a size of 384 × 256 pixels are used as the stored image group, and 10 images with a size of 80 × 80 pixels are cut out from any location in the stored image group as the target image. It was.
In addition, the partial image was cut out with a size of 64 × 64 pixels and a margin of 16 pixels in both the vertical and horizontal directions.
That is, the number of partial areas per stored image is 273, and the number of partial areas of the target image is 256.

Since the average value of RGB in a small region having a size of 8 × 8 pixels is used as the image feature, the dimension of the partial region feature is 192 dimensions.
The index assignment unit 3 and the index search unit 6 are provided with index assignment and index search based on vector quantization as in the embodiments described in “Signal Detection Method, Signal Detection Device, Recording Medium, and Program” (Japanese Patent Application Laid-Open No. 2002-236496). And the number of clusters was set to 1024.
The distance scale used was a square error, and the search threshold was 1000.

In order to confirm the effect of the present invention, first, the change in the storage capacity necessary for index holding when the margin was changed was examined.
The margin was set to the same value in both the vertical and horizontal directions.
The result of this experiment is shown in FIG. In this figure, the horizontal axis of the graph represents the margin, and the vertical axis of the graph represents the storage capacity required for indexing in bytes.
As can be seen from FIG. 12, as the margin increases, the storage capacity required for indexing is greatly reduced.

  For example, when the margin is 16, the storage capacity is 2.05 megabytes, and when the margin is 1, that is, when assigning indexes to all the partial areas, it is 124 megabytes, and 16 margins are taken. Compared to the case of 1, the storage capacity is reduced to about 1/60.

Next, we examined the change in search time (solid line) when the margin was changed.
The margin was set to the same value in both the vertical direction and the horizontal direction as in the experiment described above.
The results of this experiment are shown in FIG. In this figure, the horizontal axis of the graph represents the margin, and the vertical axis of the graph represents the search time in seconds.
As can be seen from FIG. 13, when the margin is 12, the search time is the shortest, 5.12 seconds, and when the margin is 1, the search time is 10.49 seconds, which is the longest search time. In this figure, the broken line indicates the pre-processing time before the search included in the required search time.

  In addition, a program for realizing the function of the partial image search device in FIGS. 1, 6, 8, and 10 is recorded on a computer-readable recording medium, and the program recorded on the recording medium is read by a computer system. By executing this, partial image search processing may be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer system” includes a WWW system having a homepage providing environment (or display environment). The “computer-readable recording medium” refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM, and a CD-ROM, and a storage device such as a hard disk built in the computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.

  The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

It is a block diagram which shows the structural example of the partial image search device by 1st Example of this invention. 3 is a flowchart illustrating an operation example of the partial image search device in FIG. 1. It is a conceptual diagram explaining the margin of arrangement when setting a partial image in an accumulated image. It is a conceptual diagram explaining the principle when the index search part 6 selects a cluster. It is a conceptual diagram explaining the setting of the collation window in an accumulation image. It is a block diagram which shows the structural example of the partial image search device by 2nd Example of this invention. It is a flowchart which shows the operation example of the partial image search device of FIG. It is a block diagram which shows the structural example of the partial image search device by 3rd Example of this invention. It is a flowchart which shows the operation example of the partial image search apparatus of FIG. It is a block diagram which shows the structural example of the partial image search device by the 4th Example of this invention. It is a flowchart which shows the operation example of the partial image search apparatus of FIG. It is a graph which shows the relationship between the margin in the case of setting a partial area | region at the time of searching an image, and the number of bytes of the storage capacity required for index assignment. It is a graph which shows the relationship between the margin in the case of setting a partial area | region at the time of searching for an image, and search required time.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 ... Accumulation feature extraction part 2 ... Accumulation partial area | region extraction part 3 ... Index provision part 4 ... Objective feature extraction part 5 ... Target partial area | region extraction part 6 ... Index search part 7 ... Feature collation part 8 ... Collation result determination part 9 ... Feature re-collation unit 10. Collation result re-determination unit 11... Search result candidate selection unit 12 .. Next target partial region feature selection unit 13... Next search candidate feature selection unit 14. ... Distance calculation part between target partial areas 17 ... Peripheral distance lower limit calculation part

Claims (10)

A partial image search method for detecting an image including a partial image similar to a target image to be searched from a stored image group registered in a database and a position of the partial image in the image,
An accumulated feature extraction process for extracting an accumulated feature that is a feature of each image of the accumulated image group;
An accumulation partial area feature extraction process for extracting features in the attention window at each position as accumulation partial area features while shifting the attention window of a predetermined size at a predetermined interval;
An indexing process for providing an index for labeling each of the accumulated subregion features;
A target feature extraction process for extracting a target feature that is a feature of the target image;
In the target feature, a target window of a predetermined size is set, and a target partial region feature extraction process for extracting a feature in the target window at each position as a target partial region feature while shifting the target window;
An index search process for extracting accumulated partial area features similar to the set of target partial area features as search candidate features using the index; and
A feature matching process for calculating a distance between the search candidate feature and the target partial region feature;
Based on the distance, a collation result determination process for determining whether or not there is a possibility that a partial image similar to the target image exists at the location in the accumulated image group;
In the verification result determination process, a verification window having the same size as the target image is provided for the stored feature at a location in the stored image where it is determined that there may be a partial image similar to the target image. A feature rematching process for setting and calculating a feature distance with the target feature;
Based on the feature distance, a verification result redetermination process for determining whether or not a partial image similar to the target image exists at the location in the accumulated image group;
A search result candidate selection process in which the location in the stored image determined to have a partial image similar to the target image in the matching result redetermination process is a search result candidate;
A next target subregion feature selection process for designating a target subregion feature to be matched next from the set of target subregion features;
A partial image search method comprising: a next search candidate feature selection step for designating a search candidate feature to be matched next from the set of search candidate features. In the verification result re-determination process, the feature distance is compared with a search threshold that is a threshold corresponding to the feature distance;
It is determined whether or not a partial image similar to the target image exists at the location in the stored image,
In the index search process, the accumulated sub-region feature whose distance from any target sub-region feature derived from the target sub-region feature extraction step is less than a selection threshold determined from the search threshold is used for the index. In the verification result determination process, the distance is compared with the selection threshold value, and it is determined whether or not there is a possibility that a partial image similar to the target image exists at the location in the accumulated image. The partial image search method according to claim 1, wherein: In the search result candidate selection process, locations in the stored image that are determined to have a partial image similar to the target image are selected as a predetermined number in order from the smallest feature distance to be search result candidates,
The search threshold update step of detecting a candidate based on the feature distance from the search result candidate and newly setting the distance value of the candidate as a search threshold. Image search method.   The search candidate feature is extracted in the index search process using all target partial region features derived by the target partial region feature extraction step. Partial image search method. A target representative feature extraction process for classifying the target partial region features and extracting representative features that are partial region features representing each classification;
4. The partial image search method according to claim 1, wherein, in the index search process, a representative feature derived by the target representative feature extraction process is used instead of the target partial region feature. .   In the next target partial region feature selection process, the target partial region feature extracted in the target partial region feature extraction process, in which it is detected that the distance from the accumulated partial region feature exceeds the selection threshold, is then collated. 6. The partial image search method according to claim 1, wherein collation is omitted without designating as a partial region feature to be performed. A distance calculation process between target partial areas for calculating a partial distance of each of the target partial area features;
Using the distance and the partial distance, the distance lower limit value of the distance between the accumulated partial area collated in the feature matching process and each target partial area feature extracted in the target partial area feature extraction process is calculated. And a surrounding distance lower limit calculation process.
In the next target partial area feature selection process, the distance lower limit value is compared with the selection threshold value, and a location of the target partial area feature whose distance lower limit value is lower than the selection threshold value is designated as a partial area to be collated next. The partial image search method according to claim 6. A partial image search device for detecting an image including a partial image similar to a target image to be searched from a stored image group registered in a database and a position of the partial image in the image;
A storage feature extraction unit that extracts a storage feature that is a feature of each image of the storage image group;
A storage partial region feature extraction unit that extracts a feature in the target window at each position as a storage partial region feature while shifting a target window of a predetermined size at a predetermined interval;
An indexing unit for providing an index for labeling each of the accumulated partial region features;
A target feature extraction unit that extracts a target feature that is a feature of the target image;
In the target feature, a target window of a predetermined size is set, and a target partial region feature extraction unit that extracts the feature in the target window at each position as the target partial region feature while shifting the target window;
An index search unit that extracts storage partial region features similar to the set of target partial region features as search candidate features using the index; and
A feature matching process for calculating a distance between the search candidate feature and the target partial region feature;
Based on the distance, a matching result determination unit that determines whether or not there is a possibility that a partial image similar to the target image exists in the location in the stored image group;
In the verification result determination process, a verification window having the same size as the target image is provided for the stored feature at a location in the stored image where it is determined that there may be a partial image similar to the target image. A feature re-matching unit that sets and calculates a feature distance from the target feature;
A matching result re-determination unit that determines whether or not a partial image similar to the target image exists in the portion of the accumulated image group based on the feature distance;
A search result candidate selection unit that sets a location in the stored image determined to have a partial image similar to the target image in the matching result redetermination process as a search result candidate;
A next target partial region feature selection unit for designating a target partial region feature to be matched next from the set of target partial region features;
A partial image search device comprising: a next search candidate feature selection unit that specifies a storage partial region to be collated next from the set of storage partial region features. A program for causing a computer to execute a partial image search process for detecting an image including a partial image similar to a target image to be searched from a stored image group registered in a database and a position of the partial image in the image,
An accumulated feature extraction process for extracting an accumulated feature that is a feature of each image in the accumulated image group;
A storage partial region feature extraction process for extracting features in the target window at each position as storage partial region features while shifting a target window of a predetermined size at a predetermined interval;
An indexing process for providing an index for labeling each of the accumulated partial region features;
A target feature extraction process for extracting a target feature that is a feature of the target image;
A target partial region feature extraction process 5 for setting a target window of a predetermined size in the target feature and extracting the feature in the target window at each position as a target partial region feature while shifting the target window;
Index search processing for extracting accumulated partial area features similar to the set of target partial area features as search candidate features using the index; and
A feature matching process for calculating a distance between the search candidate feature and the target partial region feature;
Based on the distance, a collation result determination process for determining whether or not there is a possibility that a partial image similar to the target image exists in the location in the accumulated image group;
In the verification result determination process, a verification window having the same size as the target image is provided for the stored feature at a location in the stored image where it is determined that there may be a partial image similar to the target image. A feature rematching process for setting and calculating a feature distance from the target feature;
Based on the feature distance, a collation result re-determination process for determining whether or not a partial image similar to the target image exists in the location in the accumulated image group;
A search result candidate selection process in which a location in the stored image determined to have a partial image similar to the target image in the matching result redetermination process is set as a search result candidate;
A next target partial region feature selection process for designating a target partial region feature to be matched next from the set of target partial region features;
A program for causing a computer to execute a partial image search process including: a next search candidate feature selection process for designating a storage partial area to be collated next from the set of storage partial area features. A computer-readable recording medium on which the partial image search program according to claim 9 is recorded.

Images