A tool based on the co-occurrence matrix to measure the performance of dynamic range reduction algorithms

Recent years have seen an increasing interest in the acquisition of scenes that have a dynamic range far greater than that of conventional acquisition devices. Some software methods have been devised to build high dynamic range (HDR) images by multi-exposure. HDR sensors have also recently appeared. On the other hand, common visualization devices still have a dynamic range that is much lower. A number of software methods have been proposed in the literature to reduce the dynamics of HDR pictures in order to fit low dynamic range displays, while preserving visual information. Usually the performance of these techniques is judged on the basis of visual results. A tool is needed for qualitative and quantitative evaluation. Common evaluation tools lack flexibility, since either they cannot express relevant features, such as local contrast (e.g. histograms), or they are little expressive for qualitative and visual evaluation (e.g. functions that measure contrast). We present a comprehensive tool, based on the co-occurrence matrix, that allows to measure many useful features and can be conveniently used for visual evaluation. We show its expressiveness examining the results of different techniques for dynamic range reduction.