Visual-Based Image Retrieval by Block Reallocation Considering Object Region

Abstract

Visual-based image retrieval based on the visual similarity over the entire image is very useful when targeting various kinds of large-volume content. This method generally divides an image into grid-shaped blocks and uses similarities based on a comparison of image features between corresponding block regions in two different images. However, the method sometimes fails in terms of object-conscious retrieval when their backgrounds are almost the same but the only object is different or object's positions and/or sizes are different. In this paper, we propose a new method featuring the reallocation of some blocks into the object region (OB-blocks) and the new similarity score with placing weight on the OB-blocks, which are derived from visual saliency map. Our proposed method could realize the "visual-based and object-conscious" image retrieval. We verified the effectiveness of this method through comparison experiments.