Combining Descriptors for Efficient Retrieval in Databases Images

Abstract

Content based image retrieval is one of the import search area. Its principle consists to search the images similar to image request by using a set of descriptors. The multimedia content description interface, MPEG-7, provides normative descriptors, such as texture, color, region and shape descriptors for effective visual content retrieval. These descriptors represent visual contents.

with numerical feature values from which the similarity could be measured quantitatively. One of the important manners of indexing images consists on extracting several visual features and afterward, using a weighted linear combination, where a weight values are assigned to the visual features. In this work, we propose to develop an efficient application based on combination of several descriptors adapted by mpeg-7.

Each image in database are described by six descriptors Angular Radial Transform, Fourier Descriptor, Curvature Scale Space, Color Structure Descriptor, Scalable Color Descriptor and EHD. Each descriptor generates a specific feature vector where the generated vector matches with a specific distance to compute the similarity between two or more images in order to found the nearest image. The application developed was tested on Mpeg7 database and Corel Database using a combination of metric to measure similarity between images and recall/precision to measure performance of search. The obtained results prove the importance and the performance of developed application.