Modeling the Spatial-chromatic Characteristics of Images by Nakagami-m Distribution

With the advent of the Internet and World Wide Web, ubiquitous multimedia information has reached every aspect of our daily lives. Multimedia data are, however, more difficult to access due to their rich and diversified interpretations. Content-based retrieval has been proposed to provide a user-friendly interface for many multimedia applications. For image retrieval, higher accuracy can be achieved by incorporating the spatial information into the classic color histogram. Unfortunately, complete description of the joint spatialchromatic feature space incurs formidable complexity of high dimensionality. To relieve the dimensionality curse, we investigate and derive the statistical model of the spatial-chromatic distribution of images. Through extensive evaluation of large image databases, we discover that a two-parameter Nakagami-m distribution well suits the purpose. The proposed simple model is confirmed by a goodness-of-fit test and a statistical similarity measure.