Reduced-reference image quality assessment based on phase information in complex wavelet domain

In this paper, we proposed a reduced-reference image quality assessment method based on the statistical information of the relative phase and the mean directional information, both of which can be obtained from the complex wavelet coefficients. Dual-tree complex wavelet transform (DT-CWT) that possesses more excellent properties than discrete wavelet transform (DWT) is used to produce complex wavelet coefficients. The basic idea underlying this algorithm is that the distribution of relative phase computed from a standard natural image possesses high consistence, while different distortions disturb the distribution in different ways. Combining with the directional information, it forms the reduced-reference features which are sensitive to most distortions in images. Finally the relationship between the difference of reduced-reference features and the subjective scores is established through a regression algorithm. The experiments are performed based on LIVE database, and the results show that the proposed method is in a leading position among the existent author-known reduced-reference image quality assessment methods.