Motion area based exposure fusion algorithm for ghost removal in high dynamic range video generation

High dynamic range (HDR) image is becoming prevalent for its superiority in recording details of the real-world scene. A conventional exposure fusion method can be utilized to generate a HDR image from a set of differently exposed low dynamic range (LDR) images according to their weight maps. However, this method causes ghosts in the HDR image if there are moving objects or the background changes in the LDR image sequence. Thus, conventional exposure fusion method is only suitable for HDR image generation, not for HDR video generation. A motion area based exposure fusion algorithm is therefore proposed to remove ghosts in HDR video generation. First, the motion areas are detected using block matching based motion estimation method. Then those areas are combined as an additional measurement to change values of original weight maps. The subjective and objective experimental results show ghost-free and highly detailed exposure fused image sequences from HDR videos.