CNN Based No‐Reference HDR Image Quality Assessment

Motivated by the problems of non-universality and over-reliance on original reference image in High dynamic range (HDR) Image quality assessment (IQA), a convolutional neural network-based algorithm for no-reference HDR is proposed. The Salience detection self-resemblance (SDSR) which extracts salient regions image, used to simulate human visual attention mechanism. Then perception network training prediction models designed according characteristics luminance contrast sensitivity. And this consists an Error estimation (Error-net), Perceptual resistance (PR-net) mixing function. experimental results indicate that method proposed has high consistency with subjective perception, value metrics Spearman rank-order correlation coefficient (SROCC), Pearson product-moment (PLCC) Root mean square error (RMSE) correspondingly reaches 0.941, 0.910 8.176 as well. It comparable classic full-reference IQA methods.