Image Indexing Based On Mpeg-7 Scalable Color Descriptor

摘要

The MPEG-7 standard defines a set of descriptors that extract low-level features such as color, texture and object shape from an image and generate metadata that represents the extracted information. In this paper we propose a new image retrieval technique for image indexing based on the MPEG-7 scalable color descriptor. We use some specifications of the scalable color descriptor (SCD) for the implementation of the color histograms. The MPEG-7 standard defines l_t - norm based matching in the SCD. But in our approach, for distance measurement, we achieve a better result by using cosine similarity coefficient for color histograms. This approach has significantly increased the accuracy of obtaining results for image retrieval. Experiments based on scalable color descriptors are illustrated. We also present the color spaces supported by the different image and video coding standards such as JPEG-2000, MPEG-1,2,4 and MPEG-7. In addition, this paper outlines the broad details of MPEG-7 Color Descriptors.