Image Retrieval Based on Dominant Color and Spatial Relationship in DCT Domain
Ling Xia, Zhi Jian Peng, Andong Cai, Mei Fa Huang · International Journal of Applied Mathematics & Statistics/International journal of applied mathematics and statistics · 2013
Nowadays images are usually represented in compressed format, most of which are compressed using the Discrete Cosine Transform (DCT). In such cases, extracting features directly from the compressed DCT domain is more efficient. Color is one of the most common visual features, but the color histogram cannot express the space distribution, thus two different color images may have the same color histogram or very similar color histograms. This paper presents a novel image retrieval scheme in the DCT domain. Firstly, dominant color blocks are selected from the image’s DCT coefficients, then, the quantized color histogram with spatial relationship is constructed. The experiment shows that the algorithm has the characteristics of less calculation, high retrieval efficiency and the ability to distinguish different color spatial relation.