Claims
- 1. A method for adaptively searching a feature vector space, the method comprising the steps of:
(a) performing a similarity measurement on a given query vector within the feature vector space; and (b) applying search conditions limited by the result of the similarity measurement obtained in the step (a) and performing a changed similarity measurement on the given query vector.
- 2. The method of claim 1, wherein the step (b) comprises the steps of:
(b-1) obtaining a plurality of candidate approximation regions by performing an approximation level filtering according to a distance measurement limited by the result of the similarity measurement obtained in the step (a); and (b-2) performing a data level filtering on said plurality of obtained candidate approximation regions.
- 3. The method of claim 2, wherein the step (a) comprises the steps of:
(a-1) obtaining a predetermined number of nearest candidate approximation regions by measuring a first plurality of distances between the query vector and each said candidate approximation region; and (a-2) obtaining a plurality of K nearest neighbor feature vectors by measuring a second plurality of distances between a plurality of feature vectors in said nearest candidate approximation regions and the query vector, where K is a positive integer.
- 4. The method of claim 3, wherein the step (b-1) comprises the steps of:
(b-1-1) calculating a K′-th shortest distance for said plurality of K nearest neighbor feature vectors obtained by said second plurality of distance measurements according to a changed distance measurement where K′ is a positive integer, and setting a calculated distance as rt+1u; and (b-1-2) calculating K′-th smallest lower bound limit for said plurality of predetermined number of nearest candidate approximation regions obtained by said first plurality of distance measurements according to said changed distance measurement and set as Φt+1u.
- 5. The method of claim 4, wherein the step (b-1) further comprises the steps of:
(b-1-3a) measuring a distance Ll(Wt+1) between said lower bound limit of at least one said nearest candidate approximation region and said query vector to determine a first new distance measurement, wherein N is a positive integer denoting the number of objects in the feature vector space and l is a variable ranging from 1 to N; (b-1-4) comparing the distance Ll (Wt+1) obtained in the step (b-1-3a) with a minimum value min (101 , rl+1u, Φt+1u) of K-th smallest upper bound limit Φ, rl+1u, and Φl+1u; wherein (b-1-5) if the distance Ll(Wt+1) is less than or equal to the minimum value mm (Φ, rt+1u, Φt+1u) setting a corresponding approximation region as a new candidate approximation reunion; and (b-1-6) if the distance Ll(Wt+1) is greater than the minimum value min (Φ, rl+1u, Φl+1u), excluding the corresponding approximation region.
- 6. The method of claim 5, wherein the step (b-1) further comprises the steps of:
(b-1-3b) measuring a distance Ul(Wt+1) between the upper bound limit of at least one said nearest candidate approximation region and the query vector for a second new distance measurement, assuming that N is a positive integer denoting the number of objects in the feature vector space and l is a variable ranging from 1 to N; (b-1-7) updating the K-th smallest upper bound limit Φ based on the distance Ul(Wt+1).
- 7. The method of claim 5, wherein the steps of (b-1-1)-(b-1-6) are repeated until the approximation level filtering is performed on all said candidate approximation regions in a database, wherein all the candidate approximation regions in said database is denoted by a positive integer (N), which represents a number of objects in said database.
- 8. The method of claim 6, wherein the steps of (b-1-1)-(b-1-6) are repeated until the approximation level filtering is performed on all said candidate approximation regions in a database, wherein all the candidate approximation regions in said database is denoted by a positive integer (N), which represents a number of objects in said database.
- 9. The method of claim 3, wherein the step (b-2) comprises the steps of:
(b-2-1) performing a third distance measurement between each of all feature vectors in said plurality of nearest candidate approximation regions and the query vector; and (b-2-2) determining K′ nearest neighbor feature vectors as retrieved vectors depending on the result of said third distance measurements performed in the step (b-2-1).
Priority Claims (1)
Number |
Date |
Country |
Kind |
00-79181 |
Dec 2000 |
KR |
|
Parent Case Info
[0001] This application is a complete application filed under 35 U.S.C § 111 (a) and claims, pursuant to 35 U.S.C. §119 (e)(1), benefit of the filing date of Provisional Application Serial No. 60/248,012 filed Nov. 14, 2000 pursuant to 35 U.S.C. § 111 (b). The Provisional Application Serial No. 60/248,012 is incorporated herein by reference. Additionally, this application claims pnority from Korean Application No. 00-79181 filed Dec. 20, 2000, which is also incorporated herein by reference.
Provisional Applications (1)
|
Number |
Date |
Country |
|
60248012 |
Nov 2000 |
US |