A new no-reference image quality assessment based on SVR fusion
Dakkar Borhen Eddine, Fella Hachouf, Zianou Ahmed Seghir · 2014
This paper presents a new concept of assessing image quality. It is based on support vector regression (SVR) fusion. Despite the variety of the proposed IQM measures, no efficient and sufficient measure gives good performance over different distortions. Motivated by this problem, a new measure for No reference Image Quality Assessment Based on SVR Fusion (NR BSVRF) is constituted. First, five recent no reference measures are selected to form a quality vector of an image, then the quality vector is fused via SVR. The SVR is trained to have a model that is used to predict the image quality. Obtained results are promising. They have shown better performance compared to existing No-reference image quality measures.