A Content Based Texture Pattern Retrieval System with Relevance Feedback

This project is designed to produce a design for a significantly new capability for storage and retrieval of texture map data. The proposed retrieval system will incorporate advanced texture classification techniques for automatic indexing of texture images through computation of image feature vectors. The system will provide for texture similarity searches utilizaing fuzzy matching of feature vectors. The user will be able to rank retrieved images according to their relevance and initiate automatic adjustment of the texture image query vector according to the relevance judgments of the samples. Relational data base search capability will be provided for storage and indexing of coding attributes for the texture images including author, data, agency, associated DFAD feature attributes and other data. Full text searchable text capability will be provided for annotation of individual texture records with algorithms descriptions, application details and other pertinent textual data. A POSIX compliant operating system will be assumed for the host which is consistent with state-of-art UNIX based simulation, mission planning, rehearsal and tactical planning workstations. An easy to learn point-and-click graphical user interface will be generated which will be compatible with the X-11 based Open Look and Motif operating environments as well as newer variants such as Windows NT.