earticle

영어

In this paper, a new method for spatial features based image retrieval is proposed. These are the features specifying location of an image segment in an image and relative spatial location between two image segments etc. Spatial feature adds extra capability of the system in addition to texture, color and shape based object recognition system. Heuristic and semantic features of the image segments are identified from low level color, texture descriptors and directly stored into the image database and easily retrieved also. Position of image segments in the image is detected and stored in the image database that will be accessed for retrieval process later. Object ontology is used to determine the spatial relation between heuristics. These spatial relations and corresponding concept are identified and stored. Spatial rank between two segments will be calculated and stored along with segment ids. Spatial rank is calculated by a new technique. Proper concepts are associated with image segment based on ontology based spatial relation.