Abstract
The layout-based image retrieval method generally uses grid-shaped division of an image into sub-blocks and similarity based on a comparison of image features between corresponding sub-block regions in two different images. However, there are some cases in which geometric differences of objects or the similarity of the background region have a negative effect. In this paper, we propose a method that uses the allocation of some sub-blocks in the object region and the similarity of highly weighted sub-blocks in the object region. We verified the effectiveness of this method through comparison experiments.
