JP2000010989A - Similar object retrieval method/device and recording medium recording similar object retrieval program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To eliminate troubles for calculating distances to all objects and to retrieve a similar object with arbitrary weight. SOLUTION: A coordinator part 16 designates the feature quantity VRi of the i-th type of a reference object, a number (f) (K, w1,... wn) designating the number of nearby objects and retrieval precision (e) on the feature quantity of i-th type and requests the retrieval of the nearby object to a nearby object retrieval part 15. The nearby object retrieval part 15 obtains (f) (K) pieces of objects, each of which is nearest to a point corresponding to a reference object in the distance of the space by using an index in the multidimensional vector space at every feature quantity type (i), and they are returned to the coordinator part 16. The coordinator part 16 gathers the returned nearby objects on all the feature quantity type and generates a candidate object group.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an object (such as an electronic museum or an electronic catalog) stored in a large number of digitized objects (images, voices, music, texts, etc.) from an object (e.g., The present invention relates to a similar object search apparatus that presents some of the top objects as search results in the order in which objects similar to the reference object) are determined to be similar.

[0002]

2. Description of the Related Art Conventionally, in a similar object search apparatus that searches for and displays an object similar to a reference object specified by a user, some feature quantities are calculated for the objects, and the similarity between the objects is calculated. The degree of similarity is calculated by calculating the distance (the distance), and the similarity is ranked. At this time, as the type of feature amount used to quantify the object, for example, when the object is an image, various types such as color tone, lightness, saturation, and the direction of a line in an image are used.

[0003] The distance is calculated as a Euclidean distance or the like between features represented by a multidimensional vector for each type of feature, and furthermore, a plurality of types of features are calculated for the same combination of objects. , The final similarity is determined in consideration of the user's viewpoint of similarity (for example, searching for similar image objects from the viewpoint of similar colors). ing. At this time, the weight of the distance of the feature amount is input by the user from the search condition input means.

In this method, the distance calculation for each type of feature is performed on all the stored objects, and each object is weighted based on the distance evaluation value for each type of feature. They used to calculate the final similarity, rank, and return the top objects.

[0005] Alternatively, there is a method in which all types of feature amounts are handled as one multidimensional vector space without dividing distance calculation for each type of feature amount, and distance calculation is performed there. In this case, the reflection of the weight, which is the user's search viewpoint, is performed by limiting the weight pattern specified by the user in advance, and taking the weight into account for each vector component of the multidimensional vector according to the pattern. It is done in such a way.

In this method, in order to speed up the search, an index is calculated in a multidimensional vector space in which weights are added in advance, and at the time of search, candidates can be narrowed down by using the index. .

[0007]

SUMMARY OF THE INVENTION The prior art as described above,
In the similar object search device based on the similarity calculation, the distance calculation has to be performed for each type of the feature amount for all the objects, so that there is a problem that the calculation cost is high.

In a method in which a predetermined pattern is weighted in advance in a multidimensional vector space of a feature value, candidates are narrowed down by adding an index, so that the calculation cost can be suppressed. However, there is a problem that an index must be provided for all the weights of the determined pattern, and the weight can be specified only for a limited number of patterns.

An object of the present invention is to solve the above-mentioned problem in the prior art, eliminate the trouble of performing distance calculation for all objects, and search for similar objects with arbitrary weights. An object search method and apparatus are provided.

[0010]

According to the similar object search method of the present invention, a feature amount of each type of stored object is calculated, and each feature amount type is stored as a point in a multidimensional vector space. In a multidimensional vector space of each feature amount type, a step of calculating an index necessary for data retrieval, a step of inputting a reference object by a user, and a step of calculating a number of similar objects desired by the user. , The step of obtaining a weight between the types of feature amounts that are search viewpoints, the step of calculating the feature amounts of the reference object for each type, the step of With respect to the distance in the multi-dimensional vector space between each of the extracted objects, a nearby object with respect to the reference object is obtained using the index. Object number is determined for each feature amount type using the weight between the feature amount types, and a neighboring object search step of obtaining objects in the vicinity by the determined number is performed. A similarity object candidate is created by summarizing all the feature amount types, and a similarity calculation step of obtaining a similarity by adding the weight to the candidate, and ordering the candidate objects according to the similarity, the user There is a step of obtaining a final result according to the designated number, and a step of presenting the final result to the user.

According to the present invention, the calculation is divided for each type of feature quantity;
Narrowing down neighboring objects using indexes in individual feature multi-dimensional vector space, distance calculation for candidates that combine nearby objects from all types of feature multi-dimensional vector space, and similarity calculation by weight Is performed.

References K. Curtis, N. Taniguchi, J. Nakagaw
a, M. Yamamuro, “A ComprehensiveImage Similarity R
etrieval System that utilizes Multiple Feature Vec
Tors in High Dimensional Space, "When searching for similar objects based on overall similarity by integrating the results of neighborhood search with multiple types of features as described in IPSJ Database System Workshop 113-28, 1997.7, The number of top searches in each feature type determines the overall search accuracy.As reported in this document, the larger the number of searches for each feature type, the higher the overall search accuracy On the other hand, how many neighbors are obtained by a search in the neighborhood search affects the time required for the search.
It is generally known that the longer the number of requests, the longer it takes. Therefore, the present invention determines the number of nearby objects to be obtained for each feature type using the weight between the feature types.

According to the second aspect of the present invention, in the neighborhood object search step, the search accuracy when finding the neighborhood object for each feature type is determined for each feature type using the weight between the feature types. Set. Here, the search accuracy refers to how close the obtained k cases are in the neighborhood search for the top k cases in descending order of similarity, and how correct the obtained k cases are based on the true top k cases calculated and obtained. It is a scale to show.

Reference S. Arya, DM Mount, NS Netany
ahu, R. Silverman, and AY Wu, “An optimal algor
ithm for approximate nearest neighbor searching,
“5th Ann. ACM-SIAM Symposium on Discrete Algorith
ms, 1994, pp.573-582. In general, reduce search accuracy (increase the value of ε in this document).
The search time is reduced.

Therefore, for a feature amount type having a small weight, the time required for calculation can be reduced as a whole without lowering the search accuracy by lowering the search accuracy.

According to a fifth aspect of the present invention, in the neighborhood object search step, the distance of the feature type of the neighborhood object in the multidimensional vector space is calculated, and the neighborhood object obtained from all the feature types is calculated. At the time when the candidate objects are collectively obtained, the distance in the multidimensional vector space of the characteristic amount type is calculated for the characteristic amount type for which the distance has not been obtained for each candidate object. Is calculated by taking the weights into consideration.

According to the sixth aspect of the present invention, in the neighborhood object search step, the distance of the feature type of the neighborhood object in the multidimensional vector space is not calculated, and the neighborhood object obtained from all the feature types is not calculated. At the same time, the distance in the multidimensional vector space of each feature amount type is calculated for each candidate object, and then, in the similarity calculation stage, the similarity is calculated by adding weights to all the candidate objects. calculate.

According to the seventh aspect of the present invention, in the neighborhood object search step, the neighborhood object searched for in the multidimensional vector space of each feature amount type is searched for in the feature amount of the type of interest at that stage. Not only the distance from the reference object but also the distance from the reference object is calculated for all types of feature amounts. However, for each object, the distance calculated once is marked so that the distance calculation does not overlap.

The case where the object is an image object will be briefly described as an example. It is assumed that hue, lightness, and saturation are used as the types of the feature amounts. Now, it is assumed that neighboring objects A, B, and C are obtained from the hue feature quantity multidimensional vector space, neighboring objects B, C, and D are obtained from the brightness space, and neighboring objects C, D, and E are obtained from the saturation space. I do. Candidate objects that summarize these are A, B, C, D, E
Becomes For each of these, three types of distances of a hue space, a lightness space, and a saturation space are obtained, and these are weighted to calculate a final similarity. , Neighboring objects A, B, and C
This means that the hue distance is calculated at the stage of obtaining a nearby object. Others are the same. “For the feature amount types for which the distance has not been determined for each candidate object,... Calculate the distance” means that for the neighboring object A, the distances of brightness and saturation have not been determined.
As for the neighboring object B, the saturation distance has not been determined yet, so this is determined. All distances have been determined for the nearby object C. Such an undetermined distance is obtained.

In claim 6, the candidate objects A, B,
That is, at the stage where C, D, and E have been obtained, all the distances have not been obtained. According to the second aspect, the distance is partially obtained, and the distance of the missing portion is obtained later. In the third aspect, the entire distance is obtained later.

In the seventh aspect, when the neighboring objects A, B, and C are determined from the hue space, the distances of all of the hues, lightness, and saturation of the neighboring objects A, B, and C are determined. Then, the candidate objects A, B, C, D,
At the stage of finding E, for all five objects,
This means that the distances of the three types of feature spaces have been determined.

A similar object search apparatus according to the present invention comprises: object data storage means for storing object data; reference object input means for allowing a user to input the reference object; Means for specifying the number of objects specified by the user, means for obtaining the weight between the types of feature amounts that are search viewpoints, and feature amount calculating means for calculating the feature amount of each type for each of the stored objects and reference objects And a feature amount storage unit for storing, among the calculated feature amounts, the feature amounts of the object stored in advance as points in a multidimensional vector space for each feature amount type; Of the feature quantities that were stored in advance, In the multidimensional vector space, an index necessary for data retrieval is calculated, an index generation / storage means for storing the index, and a reference object and each stored object are stored for each type of feature amount. With respect to the distance in the multidimensional vector space between, the number of obtaining objects in the vicinity of the reference object using the index is determined for each feature type using the weight between the feature types. A nearby object search means for obtaining only nearby objects; a group of similar objects obtained for each type of feature amount for all types of feature amounts; a candidate for a similar object; Means for calculating the similarity, and ordering and using the candidate objects according to the similarity. There has means for determining the final result to fit the specified number, and object presenting means for presenting the final results to the user.

[0023]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of a similar object search apparatus according to an embodiment of the present invention, and FIG. 2 is a flowchart showing the operation thereof.

The similar object search apparatus according to the present embodiment uses digitized object data Ob1 to Ob
m, an object data accumulating unit 10 for accumulating m, a characteristic amount calculating unit 11 for calculating characteristic amounts of the objects Ob1 to Obm, a characteristic amount / index storage unit 12 for storing the characteristic amount and an index separately generated, A reference object input unit 131 for a user to newly input a reference object from outside, a weight specification unit 132 for the user to specify a weight which is a viewpoint for measuring the similarity between objects, and a user Object number specifying unit 133 for specifying the number of objects obtained as an object, an object presenting unit 13 for presenting an object to a user
4, an interface unit 13 with the user, an index generation unit 14 for generating an index in the multidimensional vector space for each type of feature amount, and a reference object in the multidimensional vector space for each type of feature amount. On the other hand, a neighboring object search unit 15 for finding an object that is nearby at a distance in the multidimensional vector space, and upon receiving a search request from a user, divides the feature amounts of the reference object into feature amount types, The search request is delivered to the nearby object search unit 15 for each type, and the returned results are collectively referred to as candidate objects. If necessary, the distance for each feature type from the lacking reference object is obtained for each candidate object. , Calculate the similarity by taking the weight into account,
It comprises a coordinator unit 16 that returns results as similar objects by the number specified from the higher-level objects, and a temporary data storage unit 17 that temporarily stores intermediate results.

The index is an auxiliary data structure for retrieving a point at a specific position in a multidimensional space or for retrieving a point near a specified position at a high speed. Typically, there are R-tree and k-dtree. These indexes are searched by creating a hierarchical data structure depending on the position of the point and manipulating it from the source (root node). In a familiar example, meshes are cut vertically and horizontally on a road map (two-dimensional space) or the like, and this sightseeing spot is located vertically and horizontally three times.

Next, the operation of this embodiment will be described with reference to the flowchart of FIG.

The feature values Ob11, O of the object Ob1
b12,. . . , Ob1n, the feature amounts Ob21, Ob22,. . . , Ob2n,. . . , Object Obm feature amount Obm1, Obm
2,. . . , Obmn (n is the number of types of feature amounts) is calculated in advance by the feature amount calculation unit 11 and stored in the feature amount / index storage unit 12. Each object O
b1 to Obm themselves are stored in the object data storage unit 10. Further, in the multidimensional vector space of each feature amount type, the index generation unit 14 generates an index for search. This index is stored in the feature / index storage unit 12.

First, the user enters the reference object input unit 1
31 and weights w1,. . . , Wn to the object number designation unit 133
Input the number K of objects. If there is no input from the user, default values are used for the weights w1 to wn and the number K of objects (step 21). Next, the feature amounts VR1,. . . , VRn are calculated by the feature value calculation unit 11 (step 22).

Next, the coordinator unit 16 calculates the i-th type feature amount VRi of the reference object and the numbers (K and w1, w2,...) Specifying the number of neighboring objects for the i-th type feature amount. , Wn) f (K,
w1, w2,..., wn) (however, a number greater than or equal to K);
A search precision e is designated, and a request for a nearby object search is made to the nearby object search unit 15 (step 23). At this time, for example, wi × f (K) can be considered as f (K, w1, w2,..., Wn).

At this time, the search precision e differs depending on the index used. For example, the reference S. Arya,
DM Mount, NS Netanyasu, R. Silverman, and
AYWu, “An Optimal algorithm for approximate ne
arest neighbor searching, “5th Ann. ACM-SIAM Sympo
ε in sium on Discrete Algorithms, 1994, pp. 573-582. According to the present invention, for a feature amount type having a small weight, a neighborhood is searched for with reduced search accuracy. That is,
It is assumed that the value of the precision e is a value of 0 to 1, and when it is 1, it is exactly the same as the true neighborhood. In this case, e is determined in such a manner that the larger the weight wi, the larger e (closer to 1). For example, wi is weight w
When normalization is performed on the whole of l,.
It takes the value of 1, but the value of e is 0.5 + wi / 2 or sq
It is conceivable to use the value of r (wi) (where sqr (x) is the square root of x).

Neighboring object search unit 15 receiving the request
For each feature amount type i in its multidimensional vector space,
F (K) points that are close to the point corresponding to the reference object by the distance in the space are obtained using the index, and are returned to the coordinator 16 (step 24).

The coordinator unit 16 that has returned the neighboring objects collects the returned neighboring objects for all feature types and creates a candidate object group (step 25). The candidate object group is obtained, for example, as a union of neighboring objects. In this case, the number of candidate objects is maximum f (K) × n (when there is no overlap between neighboring objects for each feature amount type), and minimum f (K) (when the neighboring objects are exactly the same for each feature amount type). become.

Next, regarding the candidate object Obk,
Point V corresponding to reference object for each feature type
If the distance dki between Ri and Obki (i is the type of feature amount) has not been determined, the shortage is calculated (step 2).
6).

For each candidate object Obk, VR
1 and Obk1 are distances dk1,. . . , VRn and Obk
Once the distance dkn to n has been determined, the weight w
1,. . . , Wn are calculated to calculate the similarity (step 27). For example, w1 / dk1 + ... + wn / dkn
The similarity is calculated as follows.

Next, the candidate objects are arranged in descending order of the obtained similarity, and the top K objects are returned as a result (step 28).

When the result is returned from the coordinator section 16, the object presenting section 134 presents the result to the user (step 29).

As described above, for the feature amount type designated as having a weight of 0, the neighboring object search and the distance calculation can be omitted. Further, as the weight, not only the one specified by the user is necessarily used directly, but also a weight obtained by some method from the specified one may be used. Steps 24 and 26 can be performed in parallel with respect to the types of feature amounts.

According to another embodiment of the present invention, the distance calculation in step 26 is performed together with the step of obtaining a nearby object in step 24. In this case, step 26 is unnecessary.

Referring to FIG. 3, a similar object retrieval apparatus according to another embodiment of the present invention includes an input device 31, storage devices 32-34, an output device 35, a recording medium 36, and a data processing device 37.

The input device 31 corresponds to the reference object input unit 131, the weight specifying unit 132, and the number-of-presented-objects specifying unit 133 in FIG. The storage devices 32, 33, and 34 correspond to the object data storage unit 10, feature quantity / index storage unit 12, and temporary data storage unit 17 in FIG. 1, respectively. The output device 35 is the object presentation unit 1 in FIG.
34. The recording medium 36 is a recording medium such as a floppy disk, CD-ROM, magneto-optical disk, or semiconductor memory, and records the processing shown in FIG. 2 as a similar object search program. Data processing device 37
Is a CPU of the computer, reads a similar object search program from the recording medium 36, and executes it.

[0042]

As described above, according to the present invention, for each type of feature amount, objects near the reference object are narrowed down using the index, similar object candidates are created from these nearby objects, and the candidate Since distance calculation and calculation of similarity by weight are performed,
It is possible to search for similar objects with arbitrary weights without having to perform the distance calculation for all objects, and to reduce the number of searches for feature types that do not significantly affect the overall accuracy. By increasing the number of searches for a feature type that greatly affects the accuracy of the search, the time required for the entire search can be reduced without lowering the overall accuracy.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of a similar object search device according to an embodiment of the present invention.

FIG. 2 is a flowchart showing an operation of the similar object search device of FIG. 1;

FIG. 3 is a configuration diagram of a similar object search device according to another embodiment of the present invention.

[Explanation of symbols]

 Reference Signs List 10 object data storage unit 11 feature amount calculation unit 12 feature amount / index storage unit 13 interface unit 131 reference object input unit 132 weight designation unit 133 weighted object number designation unit 134 object presentation unit 14 index generation unit 15 neighboring object search unit 16 coordinator Unit 17 Temporary data storage unit 21 to 29 Step 31 Input device 32 to 34 Storage device 35 Output device 36 Recording medium 37 Data processing device

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5B075 ND08 ND20 ND40 NK10 NK14 NK24 NK43 PP30 PQ00 PQ46 PR06 PR10 QM08 QP10 QT03 UU40 5L096 AA02 DA01 EA23 EA31 EA39 FA15 FA69 FA76 FA81 GA08

Claims (21)

[Claims] 1. A similar object search method for extracting an object similar to a reference object, which is an object designated by a user, from a large number of accumulated objects in a descending order of similarity and presenting the same to a user. It calculates the amount of feature of each type of the accumulated object,
A stage in which each feature type is accumulated as a point in the multidimensional vector space, and in the multidimensional vector space of each feature type, an index required for data retrieval is calculated, and a reference object is used. The user inputs the number of similar objects desired as a result, the user specifies the number of similar objects desired as a result, the step of obtaining a weight between the types of feature amounts that are search viewpoints, and the feature amount of the reference object for each type. And calculating, for each type of feature amount, the distance in the multidimensional vector space between the reference object and each of the stored objects, the number of objects to be used for obtaining a nearby object with respect to the reference object using the index, A determination is made for each type of feature using the weight between the types of feature, and the determined number of nearby objects for obtaining the nearby objects is determined. A similar object candidate is created by retrieving the similar object by combining the neighboring objects obtained for each type of feature amount with respect to all feature amount types, and the similarity is calculated by adding the weight to the distance between the candidate and the reference object. A similarity calculation step of: ordering the candidate objects according to the similarity degree to obtain a final result according to a number specified by a user; and presenting the final result to a user. Object search method. 2. The method according to claim 1, wherein in the neighborhood object search step, a search accuracy for finding a neighborhood object for each feature amount type is determined for each feature amount type using a weight between the feature amount types. Similar object search method. 3. The number for finding the nearby object is:
2. The method according to claim 1, wherein the number of the feature amounts having the small weight is reduced, and the number of the feature amounts having the large weight is increased.
The described method. 4. The method according to claim 2, wherein the retrieval accuracy is reduced for the feature type having a small weight. 5. In the neighborhood object search step, a distance of each feature amount type of the neighborhood object from a reference object in a multidimensional vector space is calculated, and the neighborhood objects obtained from all feature amount types are collectively determined as candidates. At the time when the object is obtained, the distance from the reference object in the multidimensional vector space of the characteristic amount for which the distance has not been determined for each candidate object is calculated. The similar object search method according to claim 1, wherein the similarity is calculated by adding a weight to a distance between the candidate object and the reference object. 6. The neighborhood object search step does not calculate the distance of each feature amount type of the neighborhood object from the reference object in the multidimensional vector space, and collects the neighborhood objects obtained from all the feature amount types. At the time when the candidate object is obtained, the distance in the multidimensional vector space of each feature amount type is calculated for each candidate object, and then, in the similarity calculation step, the weights are added to the distances from all the candidate objects to the reference object. The similar object search method according to any one of claims 1 to 4, wherein the similarity is calculated by using: 7. In the neighborhood object search step, for the neighborhood object searched in the multidimensional vector space of each feature amount type, only the distance from the reference object in the feature amount of the type of interest at that stage. 5. The method according to claim 1, wherein distances from the reference object are calculated for all types of feature amounts, but, for each object, the distance calculated once is marked so that the distance calculation does not overlap. 2. The similar object search method according to claim 1. 8. A similar object search apparatus that extracts objects similar to a reference object, which is an object designated by a user, from a large number of accumulated objects in a descending order of similarity and presents them to the user. Object data storage means for storing object data; reference object input means for a user to input the reference object; and object number specification means for the user to specify the number of similar objects desired as a result. A means for acquiring a weight between types of feature amounts that are search viewpoints; a feature amount calculating unit for calculating a feature amount for each type for each of the stored objects and reference objects; and the calculated feature amounts. Of the features of the objects stored in advance, A feature amount storage unit for storing as a point in a multidimensional vector space for each feature type, and a multidimensional feature of each feature type for a feature amount of an object stored in advance among the calculated feature amounts. In the vector space, an index necessary for data retrieval is calculated, an index generation / storage means for storing the index, and, for each type of feature amount, an index between the reference object and each stored object. With respect to the distance in the multidimensional vector space, the number of objects to be searched for in the vicinity of the reference object using the index is determined for each feature type using the weight between the feature types. Object retrieving means, summarizes nearby objects obtained for each feature type for all feature types Make candidates of similar objects, the similarity calculation means for calculating the similarity in consideration of the weight for that candidate, in accordance with the degree of similarity, and orders the candidate object,
A similar object search device comprising: means for obtaining a final result according to a number designated by a user; and object presenting means for presenting the final result to a user. 9. The method according to claim 8, wherein the neighborhood object search unit determines the search accuracy when finding a neighborhood object for each feature amount type by using a weight between the feature amount types. Similar object search device. 10. The method according to claim 8, wherein the neighboring object search unit decreases the number of the objects to be searched for the feature type having the small weight and increases the feature number for the feature type having the large weight. apparatus. 11. The apparatus according to claim 9, wherein said neighborhood object search means lowers search accuracy for the feature amount type having the smaller weight. 12. The neighborhood object search means calculates a distance of each feature amount type of the neighborhood object from a reference object in a multidimensional vector space, and collectively determines a neighborhood object obtained from all feature amount types as a candidate. At the time when the object is obtained, the distance from the reference object in the multi-dimensional vector space of the characteristic amount type for which the distance has not been determined for each candidate object is calculated. The similar object search device according to any one of claims 8 to 11, wherein the similarity is calculated by adding a weight to a distance between the candidate object and the reference object. 13. The neighborhood object search means does not calculate the distance of each feature type of the neighborhood object from the reference object in the multidimensional vector space, and collects the neighborhood objects obtained from all the feature types. At the time when the candidate object is obtained, the distance in the multidimensional vector space of each feature amount type is calculated for each candidate object, and thereafter, the similarity calculating means takes into account the weight of the distance to the reference object for all the candidate objects. The similar object search device according to any one of claims 8 to 11, wherein the similarity is calculated by using: 14. The neighborhood object means may determine not only the distance from a reference object in a feature amount of a type of interest at that stage, but also the distance of a searched neighborhood object in a multidimensional vector space of each feature amount type. , Calculate the distance from the reference object for all kinds of features, but mark the distance calculation once for each object so that the distance calculation does not overlap,
A similar object search device according to any one of claims 8 to 11. 15. A similar object search program for extracting an object similar to a reference object, which is an object designated by a user, from a large number of stored objects in descending order of similarity and presenting the object to a user. It calculates the amount of feature of each type of the accumulated object,
A process for storing the points required in the multidimensional vector space for each feature type and calculating the indexes required for data search in the multidimensional vector space for each feature type, Calculating the feature amount of the reference object obtained for each type, and for each type of feature amount, regarding the distance in the multidimensional vector space between the reference object and each of the stored objects, the reference object using the index. The number of nearby objects to be obtained is determined for each feature amount type using the weight between the feature amount types, and a nearby object search process for obtaining the determined number of nearby objects by the determined number is performed. A candidate for a similar object is created by combining the obtained neighboring objects for all feature types, and the candidate A similarity calculation process of calculating a similarity by adding a weight to a distance to a reference object, and a process of ordering the candidate objects according to the similarity and obtaining a final result according to a number specified by a user; A recording medium recording a similar object search program for causing a computer to execute the process of presenting the final result to a user. 16. The method according to claim 15, wherein in the nearby object search processing, a search accuracy for obtaining a nearby object for each feature amount type is determined for each feature amount type using a weight between the feature amount types. Recording medium. 17. The neighborhood object search process according to claim 15, wherein the number of the neighborhood objects to be obtained is reduced for the feature amount type with the small weight and increased for the feature amount type with the large weight. recoding media. 18. The recording medium according to claim 16, wherein said neighborhood object search processing lowers search accuracy for the feature amount type having the smaller weight. 19. The neighborhood object search processing calculates a distance of each feature amount type of the neighborhood object from a reference object in a multidimensional vector space, and collectively determines a neighborhood object obtained from all feature amount types as a candidate. At the time when the object is obtained, the distance from the reference object in the multidimensional vector space of the type of the characteristic amount for which the distance has not been obtained for each candidate object is calculated. 19. The recording medium according to claim 15, wherein the similarity is calculated by adding a weight to a distance between the candidate object and the reference object. 20. The neighborhood object search processing does not calculate the distance of each feature type of the neighborhood object from the reference object in the multidimensional vector space, but collects the neighborhood objects obtained from all the feature types. At the time when the candidate object is obtained, the distance in the multidimensional vector space of each feature amount type is calculated for each candidate object, and then, in the similarity calculation step, in the similarity calculation step, all the candidate objects are compared with the reference object. 19. The recording medium according to claim 15, wherein a similarity is calculated by adding a weight to the distance. 21. The neighborhood object search processing is performed on the neighborhood object searched for in the multidimensional vector space of each feature amount type by using only the distance from the reference object by the feature amount of the type of interest at that stage. 19. The method according to claim 15, wherein distances from the reference object are calculated for all types of feature amounts, but, for each object, the distance calculated once is marked so that the distance calculation does not overlap. 2. The similar object search method according to claim 1.