Feature Analysis of Sketch Based Image Retrieval System

Abstract

Although many researches are increased in Sketch Based Image Retrieval (SBIR) field, it is still difficult to bridge the gap between image and sketch matching problem. Feature extraction is the critical role for SBIR to get efficient matching. In this paper, the proposed feature descriptor called Edge Orientation Histogram (EOH) for sketch based image retrieval (SBIR) system is presented. The features of database images and query sketch are extracted by EOH descriptor. And then cosine similarity measure is applied for matching features. The retrieved images similar with query sketch are displayed by rank order. Mean Average Precision (MAP) is measured as evaluation criteria. The Flickr15K benchmark dataset is used to evaluate the performance of this system.