Image quality assessment by using neural networks

This paper presents a model using neural networks for image quality assessment. The proposed system aims at evaluating the difference in the perceived quality when a static image is processed with an enhancement algorithm. A CBP neural network is designed to mimic the human perception. Objective features are worked out on a block-by-block basis from both the original and the enhanced image; they feed the neural network, which yields as output the quality rating. Experimental results confirm the approach validity, as the system provides a satisfactory approximation of subjective opinions.