Quantization of Colors Using Median of Pixels for Color Correlogram

摘要

Content based image retrieval (CBIR) mostly uses color correlogram for image features because it extracts not only the color distribution of pixels in images like color histogram, but also extracts the spatial information of pixels in the images. The size of color correlogram depends on the number of quantized colors used for feature extractions. Usually, 64 quantized colors are used, and hence the size of the correlogram is 64×64 for unit distance. In this paper, we reduce the size of color correlogram to 9×9 by quantizing the color pixels using the median of pixels within a small 3×3 block. The proposed algorithm has smaller size of the correlogram and gives comparable image retrieval results.