A Method for Image Indexing using Entropy and Color Correlation features

摘要

To improve the discrimination power of color indexing techniques, we propose a new indexing algorithm that divides an image into higher and lower entropy regions, and then, that calculates the color correlation features between each type of regions. Our indexing method proposed in this paper has two important characteristics. One is a local entropy feature. An image is divided into two regions, lower and higher entropy regions, in order to avoid the problems caused by general global features. Here, lower and higher regions are classified by an entropy feature. The other is a Color Correlation Feature. We used 2-dimensional probability distribution functions of an image to obtain color moment features, and thus, obtain more information using 2-dimensional probability distribution function(PDF) than using 1-dimesional probability distribution functions (i.e. histogram).