Abstract:

Image retrieval based on texture in discrete cosine transform （DCT） compressed domain is proposed by introducing the spatial information of DCT blocks and taking the DCT coefficients into the feature extraction. Firstly, the definition of complex is introduced. Secondly, a complex histogram is presented to extract the texture information. Meanwhile, the effect of the DCT coefficients on the retrieval precision is discussed, the weighting values for histograms are presented. The statistical distribution of DCT blocks and the spatial distribution of DCT coefficients are taken into account as well, thus avoiding the mistaken retrieval and losing retrieval. Experiments show that the new method will show a good performance both in retrieval efficiency and effectiveness.

Key words: content-based image retrieval, discrete cosine transform , weighted complex histogram