In this work, we propose an approach for description, indexing and retrieval in large shapes databases. We introduce two contributions: the first one consists to propose a new shape descriptor which is invariant to translation, scale and rotation. This latter, is a multi-scale version of the Berretti et al. shape descriptor, which has the main drawback to not satisfy the rotation invariant criterion. In the second contribution, we propose to organize the indexes set of all shapes in the database using single index structure named M-tree, this tree index structure will be associated to all the shapes in the database instead to each one. By using a large shapes database, the obtained results highlight the effectiveness of our approach.