Ghost-Free High Dynamic Range Imaging via Moving Objects Detection and Extension

High dynamic range imaging(HDRI) techniques are proposed to synthesize high dynamic range (HDR) images from multi-exposure images. However, ghost artifacts may appear if images are synthesized directly when there are moving objects in the scene. This paper presents an algorithm to composite a HDR image from multi-exposure images without ghost artifacts. To remove the ghosts in the final image, the proposed algorithm firstly produces a 0-1 map based on a Markov Random Field(MRF) framework. The moving areas are detected and marked with 1. Then, moving areas are extended and used in the final exposure fusion step. The marked pixels are assigned zero weights to prevent ghost artifacts.