MRF-based high dynamic range image generation

Image enhancement using high-dynamic range (HDR) images is widely exploited; however, it is limited by detail loss and excessive color generation. In addition, capturing HDR images by commercial digital cameras is problematic. In this paper, we propose an image enhancement technique of fusing two images with different exposures. In order to reduce unnatural color changes in the fused image, initially we modify the lightness of the less-exposed image according to that of the highly exposed image. Then, we design a Markov random field model (MRF) by considering a gradient, chrominance, and smoothness constraint. Further, the MRF model is optimized via belief propagation. Experimental results show that the proposed algorithm generates more natural results than other state-of-the-art algorithms.