Generalized Average Histogram Based Image Retrieval
Zhu Lixia · Bulletin of Science and Technology · 2009
Conventional color histogram approaches to image retrieval lack the expressions of spatial information and relativity between adjacent pixels.In this paper,we introduce Generalized Image(GI),which combines the original image with its smoothed image to form a binary set,to mitigate this problem.And we present a new retrieval method based on generalized average color histogram.In the algorithm,color quantifier is extended to GI,a feature vector of 143 dimensions is get successfully also.After computing the cluster centers for each dimension and counting the number of combination number,an improvement on the conventional color histogram called generalized average color histogram(GACI) is proposed.Moreover,for better multicolor matching,the 143 dimensions are arranged according to pixel numbers of each bin,then index images by the references color set algorithm.Compared with the typical histogram method,our new method in general achieves more accurate and efficiency.