Relevance feedback for sketch retrieval based on linear programming classification

Abstract

Relevance feedback plays as an important role in sketch retrieval as it does in existing content-based retrieval. This paper presents a method of relevance feedback for sketch retrieval by means of Linear Programming (LP) classification. A LP classifier is designed to do online training and feature selection simultaneously. Combined with feature selection, it can select a set of user-sensitive features and perform classification well facing a small number of training samples. Experiments prove the proposed method both effective and efficient for relevance feedback in sketch retrieval. © Springer-Verlag Berlin Heidelberg 2006.