Towards a blind image quality evaluator using multi-scale second-order statistics
作者：
Highlights：
• Second-order statistics are of great value in image quality prediction.
• The statistical regularities between sub-band coefficients in the wavelet domain can be used to quantify image degradation.
• Bivariate generalized Gaussian distribution can be utilized to model the spatially adjacent bandpass responses.
• The histogram of Gaussian derivative pattern in the spatial domain can be employed to capture image structural degradation.
• Features extracted in both wavelet domain and spatial domain are of great importance in image quality prediction.
摘要
•Second-order statistics are of great value in image quality prediction.•The statistical regularities between sub-band coefficients in the wavelet domain can be used to quantify image degradation.•Bivariate generalized Gaussian distribution can be utilized to model the spatially adjacent bandpass responses.•The histogram of Gaussian derivative pattern in the spatial domain can be employed to capture image structural degradation.•Features extracted in both wavelet domain and spatial domain are of great importance in image quality prediction.
论文关键词：Image quality assessment,Bivariate statistics,Derivative pattern,No-reference
论文评审过程：Received 5 September 2018, Revised 17 October 2018, Accepted 20 November 2018, Available online 23 November 2018, Version of Record 29 November 2018.
论文官网地址：https://doi.org/10.1016/j.image.2018.11.003
