VISUAL IMAGE RETRIEVAL BY ELASTIC MATCHING OF USER SKETCHES

Effective image retrieval by content from database requires that visua l image properties are used instead of textual labels to properly inde x and recover pictorial data. Retrieval by shape similarity, given a u ser-sketched template is particularly challenging, owing to the diffic ulty to derive a similarity measure that closely conforms to the commo n perception of similarity by humans. In this paper, we present a tech nique which is based on elastic matching of sketched templates over th e shapes in the images to evaluate similarity ranks. The degree of mat ching achieved and the elastic deformation energy spent by the sketch to achieve such a match are used to derive a measure of similarity bet ween the sketch and the images in the database and to rank images to b e displayed. The elastic matching is integrated with arrangements to p rovide scale invariance and take into account spatial relationships be tween objects in multi-object queries. Examples from a prototype syste m are expounded with considerations about the effectiveness of the app roach and comparative performance analysis.