Sketch-based retrieval in large-scale image database via position-aware Silhouette Matching

Abstract

We propose an interactive sketching tool called SKIT to explore image database. The aim is to achieve fast result convergence according to the visual user query. Our main contribution is a new interactive image exploration approach which dynamically adapts to user sketches and provides feedback. The novel user interface is suitable for a range of interactive image-database access applications. In addition, we propose a position-aware matching approach for SKIT to support translation-free sketch searching. Experimental results demonstrate that our method outperforms state-of-the-art approaches with respect to the superior user interface and matching results.