Method and system for indexing and searching high-dimensional data using signature file

A method for searching high-dimensional data using a signature file in a filtering-based indexing scheme, the method comprising:

• receiving searching input information, including a query feature vector, a feature vector file identifier, a first-level signature file identifier, and a second-level signature file identifier;
• calculating a first-level query signature and a second-level query signature by using the query feature vector;
• calculating a similarity between a second-level signature file and the second-level query signature;
• performing a first filtering operation to obtain a primary candidate cell group containing one or more signatures having a high similarity to the second-level query signature;
• calculating a similarity between a first-level signature file and the first-level query signature by using first-level signature entry identifier information contained in the primary candidate cell group;
• performing a secondary filtering operation to obtain a secondary candidate cell group containing one or more signatures having a high similarity to the first-level query signature;
• and
• obtaining a searching result by calculating a similarity between an entry feature vector and a query feature vector of the secondary candidate cell group, and returning the obtained searched result.

A method for calculating a corrected query signature (Sic) in a filtering-based indexing scheme, the method comprising:

• (a) calculating a correction value (cv);
• (b) calculating a first signature (U(Si)) where the correction value (cv) is added, and a second signature (L(Si)) where the correction value (cv) is subtracted;
• (c) comparing the first signature (U(Si)) with the second signature (L(Si));
• and
• (d) setting the first signature (U(Si)) as a corrected signature value when the first signature (U(Si)) and the second signature (L(Si)) are equal to each other, and setting the first signature (U(Si)) and the second signature (L(Si)) as a corrected signature value when the first signature (U(Si)) and the second signature (L(Si)) are different from each other.

A system for searching high-dimensional data using a signature file in a filtering-based indexing system, the system comprising:

• a feature vector extracting means for extracting a feature vector from a multimedia object;
• an object storing means for storing the multimedia object in a database and converting an object identifier;
• an indexing generating means for generating an index using the feature vector extracted by the feature vector extracting means;
• a searching means for searching the stored object by using a query feature vector that is extracted from the multimedia object by the feature vector extracting means;
• and
• a storing means for storing the multimedia object received from the object storing means, and searching an object in cooperation with the index generating means and the searching means.