PIQI: perceptual image quality index based on ensemble of Gaussian process regression

Digital images contain a lot of redundancies, therefore, compression techniques are applied to reduce the image size without loss reasonable quality. Same become more prominent in case videos which contains sequences and higher ratios achieved low throughput networks. Assessment quality such scenarios has particular interest. Subjective evaluation most is infeasible so objective preferred. Among three measures, full-reference reduced-reference methods require an original some form calculate unfeasible as broadcasting, acquisition or enhancement. Therefore, no-reference Perceptual Image Quality Index (PIQI) proposed this paper assess digital calculates luminance gradient statistics along with mean subtracted contrast normalized products multiple scales color spaces. These extracted features provided stacked ensemble Gaussian Process Regression (GPR) perform perceptual evaluation. The performance PIQI checked on six benchmark databases compared twelve state-of-the-art competitive results achieved. comparison made based RMSE, Pearson Spearman correlation coefficients between ground truth predicted scores. scores 0.0552, 0.9802 0.9776 respectively for these metrics CSIQ database. Two cross-dataset experiments performed check generalization PIQI.