An image retrieval algorithm using multiple query images

In this paper, an image retrieval algorithm using multiple query images is proposed. The algorithm is based on multi-histogram intersection techniques. For each query image, a color histogram and a texture histogram are extracted. Then multi-histogram intersection is used to measure the similarity between the query images and each image in the database. The ranking in similarity is used to determine the images to be retrieved. This approach can be applied to image retrieval with relevance feedback and to component based image retrieval. Results are provided showing the improvement in precision that is afforded by the multi-example retrieval paradigm.