Effect of Contrast Measures on the Performance of No-Reference Image Quality Assessment Algorithm for Contrast-Distorted Images

Most of the no-reference image quality assessment for contrast distorted images (NR-IQA-CDI) algorithms use global standard deviation as a measure for contrast. On the other hand, Michelson and Weber contrast measures - compared to the standard deviation - have lesser computational complexity, and could be considered as potential substitutes if they do not degrade the performance of the NR-IQA-CDI algorithm significantly. In this regard, this paper investigates the effect of substituting the standard deviation with Michelson or Weber contrast measures to find out if the NR-IQA-CDI algorithm could be improved in terms of its computational complexity. The obtained results show that both Michelson and Weber contrast measures, significantly, enhance the performance of NR-IQA-CDI. Consequently, they can easily replace the standard deviation. Moreover, the global Weber contrast measure is found to be the best alternative for the standard deviation since it shows the best improvement in both the overall prediction accuracy and computational complexity.