Evaluation of degraded images using adaptive Jensen-Shannon divergence

This paper investigates the properties of the Jensen-Shannon divergence for visual quality assessment. It will be shown that Jensen-Shannon divergence is a potential adaptive tool thanks to a good correlation with human perception, a reduced computational effort and a limited amount of information on the original image. In addition, it gives a score to visual quality in terms of bits and it has theoretical relationships with some well known and highly performing image quality assessment measures. Preliminary experimental results show the potential of the proposed measure in predicting the amount of degradation in the analysed image. This is a valid feature to successfully use in the preservation and restoration of archived material.