An Efficient Image Retrieval Approach Based on Spatial Correlation of the Extrema Points of the IMFs

In this paper we propose a new content-based image retrieval approach based on spatial correlation of extrema points of the IMFs. This consists, first, in decomposing the image by the BEMD into its IMFs and, then, constructs the descriptor vector from the spatial correlation of the extrema points of each IMF. The choice of the BEMD is based on two major reasons: the first is that the obtained IMFs can be seen as a filter bank results like in wavelet decomposition, and the second is that the spatial information of the decomposed image is retained within the IMFs (each IMF contains information of a specific scale). The efficiency, the stability and the robustness of the proposed approach are tested on two different image databases constituted by images with different orientations and different scale transformation. The simulation results are very satisfactory compared with some existent shape descriptors like Gabor and Wavelet-Db2, and it shows the interest of these descriptors in handing image transformations such as image rotation and image scaling. This new approach allows a finer research by exploring the modes in the increase order.