Shape-Based Visual Query Rewriting

Abstract

A visual query is based on pictorial representation of conceptual entities and operations. One of the most important features used in visual queries is the shape. Despite its intuitive writing, a shape-based visual query usually suffers of a complexity processing related to two major parameters: 1-the imprecise user request due to the user certainty or to the vagueness of his need, 2-shapes may undergo several types of transformation like occlusion, rotation, etc. that need to be considered. This is why several methods are provided in the literature to assist the user during query writing. On one hand, relevance feedback technique is widely used to rewrite the initial user query. On the other hand, shape transformations are considered by current shape-based retrieval approaches without any user intervention. In this paper, we present a new cooperative approach based on the shape neighbourhood concept allowing the user to rewrite a shape-based visual query according to his preferences with high flexibility in terms of including (or excluding) only some shape transformations and of result sorting.