Fast global tone mapping for high dynamic range compression

High dynamic range (HDR) images cannot be displayed directly on traditional low dynamic range (LDR) devices. Therefore, tone mapping technology is developed to nonlinearly map HDR images to LDR ones. As a universally acknowledged global tone mapping operator (TMO), Reinhard algorithm practically show low computational efficiency, thus limiting its real-time applications. This paper has unveiled the efficiency bottleneck of the algorithm that the percentage filtering and color gamma correction dominate most of the running time. Accordingly, we propose a quick k-th selection algorithm for screening top 1% maximum and minimum pixels from the HDR image. Further, we present a fast gamma correction based on look-up table, which takes place of the time-consuming power function. The experimental results show that our optimization approach improves the operation efficiency of Reinhard TMO by 9 times without reducing the accuracy.