Windsurf: region-based image retrieval using wavelets

In this paper we present WINDSURF (Wavelet-Based Indexing of Images Using Region Fragmentation), a new approach to content-based image retrieval. The method uses the wavelet transform to extract color and texture features from an image and applies a clustering technique to partition the image into a set of "homogeneous" regions. Similarity between images is assessed by using the Bhattacharyya distance to compare region descriptors, and then combining the results at image level. Experimental results on a testbed of 10000 general-purpose images show that our approach is very effective in retrieving images that are "semantically" similar to the query image. In particular, we compared results of WINDSURF with the approach by Stricker and Orengo, showing that that a significant improvement is obtained in the quality of the result.