The article describes a method for a parameter tuning for a family of image quality assessment methods based on the concept of SSIM. The method employs curve fitting to model SSIM-MOS relationship then uses inverse relationship to calculate intended measure values and finally the log-log regressive model to estimate parameters for components constituting the measure. The regression was implemented by minimizing of least squares (L2) and least absolute deviation (L1). The results were verified against well tested reference databases.
展开▼