Multi-indicator image quality assessment of smartphone camera based on human subjective behavior and perception

As smartphones are widely used in daily lives, manufacturers and customers are increasingly concerned about the performance of smartphone cameras. However, due to the unique distortion types, there are relatively few image quality assessment (IQA) methods for smartphone images. In this paper, we propose a smartphone photo quality assessment model, which scores from four quality aspects: color, texture, noise and exposure. Based on human observation behaviors of different indicators, two novel image cropping methods viz. SalGAN-crop based on saliency prediction and SSIM-crop based on structural similarity are proposed. Different features are after-wards extracted by simulating human subjective perception, and the predicted scores are finally given by AdaBoost regression analysis. Experimental results reveal that our model can provide more accurate scores than traditional methods.