Opinion-Unaware No-Reference Image Quality Assessment of Smartphone Camera Images Based on Aesthetics and Human Perception

In this work, we evaluate the quality of images taken by smartphone cameras. By extracting localized features from different contents, images can be analyzed in terms of exposure, noise, color and texture. Coarse-grained method, which simulates the Ising model system, is utilized to evaluate the exposure performance. Gabor transform and Tamura texture features, which imitate the human visual perception, are used to grade noise and texture. For color, hue histograms are adopted. Comparing our computational result with the ranking given by experts, it is found that the averaged SROCC values are greater than 0.35.