Reduced-Reference Image Quality Assessment Based on Average Directional Information

In this paper, we introduce a new approach for general purpose reduced-reference image quality assessment (RRIQA) based on the average directional information (ADI), which is obtained from complex wavelet coefficients. The dual-tree complex wavelet transform (DT-CWT) possessing more excellent properties than the discrete wavelet transform (DWT) such as shift invariance and multi-directional selectivity is used to produce complex wavelet coefficients. Then the inter-coefficient product (ICP) is computed to represent the feature orientation (through the complex phase) and the feature strength (through the complex magnitude). We consider the distortion progress affect the directional features in the reference image, so the difference between the ADI of the distorted image and the reference image can be an indicator that quantizes the degree of degradation. The experiments are performed based on LIVE database, and the results show that the proposed approach is comparable to PSNR and some of the state-of-the-art RR approaches, meanwhile has relatively small size of RR features.