Regional Feature Covariance Measure based Efficient Color Image Classification using Color Values Vennila V. Baby*, Dr Gnanamurthy R. K.** *Research Scholar, Anna University, Chennai, India **Principal, S. K. P. Engineering College, Tamil Nadu, India Online published on 15 September, 2016. Abstract Extracting similar images from the multimedia data set has been approached in different ways. The methods use various measures in identifying similar color values, but produces higher false classification. To overcome the deficiency in color image classification and to produce higher precision in results, a regional feature covariance measure based approach is discussed in this paper. The method splits the entire color image into number of sub sampling images. For each subsample image, the method generates a regional covariance matrix. First, the method generates histogram of the region according to the color values of the pixels. Similarly, for each region or sub sample image and histogram is generated. Then the method computes the covariance between the histogram values. Then we compute the covariance measure between each neighbor region obtained. Using the regional covariance matrix and covariance measure computed, the method generates the feature distribution vector array. The feature distribution vector array has much entry for different color values which contains the information about number of regions a color value present, the impact of distribution and so on. Using the feature distribution vector, the method computes the similarity value to identify similar values. The method improves the performance of color image classification and produces higher precision in classification. Top Keywords Color Images, Image Classification, Regional Features, Covariance Measure, Feature Distribution. Top |