In this paper, we present an original and new framework to allow complex multiple regions queries based on composition of regions categories. This problematic has merely not been addressed in the litterature even though it corresponds to photo-agency usage scenario. We first present the image segmentation technique specifically designed for image retrieval by regions based on the clustering of Local Distributions of Quantized Colors (LDQC). Then, categories of similar regions are formed automatically by clustering their visual features. The representative regions for each category give to the user an overview of all the regions. They allow him to query the database by specifying different types of regions which should appear in the retrieved images but also types which shouldn't. This enables our system to retrieve images from complex queries of region categories compositio- n. This query mode differs completely from the classic Query-by-Example paradigm since there is no query image. user has a clear idea of what he is looking for. allows him to query Our retrieval system handles region range query by considering regions from the same category but also from neighbor categories. The resulting indexing and retrieval implementation turns out to be simple and very fast even on very complex query compositions and large image database. It was tested on a database of 9,995 images from the Corel image database.