Partial image retrieval using SIFT based on illumination invariant features

Abstract

In content-based image retrieval (CBIR), the apparent color of objects are strongly influenced by the illumination variation, and this may affect retrieval results adversely. In this work, we propose a framework which can find partial object matchings by using illumination invariant local features, and have achieved image retrieval robust to apparent color changes. Additionally, partial similarities to query images such as logos and trademarks under different illumination conditions were sought, and the system's robustness to apparent color changes were verified.