Image Retrieval Based on DCT Coefficients and Color Histogram
Lixia Xie · Journal of Shaoyang University · 2015
In order to describe shape,texture and color features,this paper presents an image retrieval algorithm based on DCT coefficients statistical characteristics and color histogram. Firstly,quantize the pixel values into 64 colors,and construct color histogram to describe color feature. Secondly,divide the image into 8×8 sub-block,perform DCT on each sub-block. Lastly,according to the DCT coefficients of each block to obtain the two statistical characteristics,one is to calculate the mean and variance of each sub block DCT coefficient,and construct the mean-variance histogram to describe texture feature; the other is to construct AC coefficient difference histogram by using those front 9 AC coefficients so as to describe the local shape feature. A new image retrieval feature vector is formed by combining the color histogram with the mean-variance histogram and the AC coefficient difference histogram.Experimental results indicate that the algorithm can describe the shape,texture and color features and has higher precision and recall rate,in all,it has a better performance than other image retrieval methods.