No-reference image quality assessment based on log-derivative statistics of natural scenes

摘要: We propose an efficient blind/no-reference image quality assessment algorithm using a log-derivative statistical model of natural scenes. Our method, called DErivative Statistics-based QUality Evaluator (DESIQUE), extracts quality-related features at two scales in both the spatial and frequency domains. In domain, normalized pixel values are modeled ways: pointwise-based statistics for single pairwise-based relationship pairs. log-Gabor filters used to extract fine image, which also by statistics. All these can be fitted generalized Gaussian distribution model, estimated parameters fed into combined frameworks estimate quality. train our models on LIVE database optimized support vector machine learning. Experiment results tested other databases show that proposed not only yields substantial improvement predictive performance as compared state-of-the-art no-reference methods, but maintains high computational efficiency.