Human visual system (HVS) can visualize all the brightness levels of the scene through visual adaptation. However, the dynamic range of most commercial digital cameras and display devices are smaller than the dynamic range of human eye. This implies low dynamic range (LDR) images captured by normal digital camera may lose image details. We propose an efficient approach to high dynamic (HDR) image fusion that copes with image displacement and image blur degradation in a computationally efficient manner, which is suitable for implementation on mobile devices. The various image registration algorithms proposed in the previous literatures are unable to meet the efficiency and performance requirements in the application of mobile devices. In this paper, we selected Oriented Brief (ORB) detector to extract local image structures. The descriptor selected in multi-exposure image fusion algorithm has to be fast and robust to illumination variations and geometric deformations. ORB descriptor is the best candidate in our algorithm. Further, we perform an improved RANdom Sample Consensus (RANSAC) algorithm to reject incorrect matches. For the fusion of images, a new approach based on Stationary Wavelet Transform (SWT) is used. The experimental results demonstrate that the proposed algorithm generates high quality images at low computational cost. Comparisons with a number of other feature matching methods show that our method gets better performance.
© (2014) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Yan Liu and Feihong Yu
A fast and automatic fusion algorithm for unregistered multi-exposure image sequence
", Proc. SPIE 9216, Optics and Photonics for Information Processing VIII, 921613 (September 19, 2014); doi:10.1117/12.2060036; http://dx.doi.org/10.1117/12.2060036