Improving the robustness to image scale of the Total Variation of Difference metric
Marius Pedersen, Ivar Farup · 2016
Objective image quality assessment has received a lot of attention in the last decades, and it is still an unsolved challenge. One of the problems with many existing image quality metrics is that they suffer from scale differences, i.e. images have been rated similar by observers but according to the image quality metrics the images are different. We propose a normalization step as a solution to this problem for one of the state-of-the-art metrics, the Total Variation of Difference (TVD) metric. The normalization is similar to Michelson contrast, and experimental results show that the proposed normalization significantly increases the performance of the TVD metric.