Description

HDRI estimation based on the Markov random field (MRF) model. We can construct the HDRI by taking into consideration displacements, underexposure and overexposure, and occlusions. The displacement vectors, as well as the occlusion and the saturation, are detected by the MAP estimation. In our method, we do not need to estimate accurate motion vectors but displacement to the pixel with the closest irradiance, whereas the conventional methods such as try to accurately estimate the motion. This relaxation improves the final quality of the HDRI. The occlusion and the saturation are clearly classified and then separately treated, which results in the accurate removal of ghosting artifacts. During this process, displacements of the images caused by object movements often yield motion blur and ghosting artifacts. To address the problem, this paper presents an efficient and accurate multiple exposure fusion technique for the HDRI acquisition. Our method simultaneously estimates displacements and occlusion and saturation regions by using maximum a posteriori estimation and constructs motion-blur-free HDRIs. We also propose a new weighting scheme for the multiple image fusion. We demonstrate that our HDRI acquisition algorithm is accurate, even for images with large motion.