Photometric linear relation based ghost-free exposure fusion approach

Exposure fusion is a high dynamic range (HDR) image generation method for directly fusing multi-exposure images of the same static scene. However, if the captured scene is dynamic with moving objects, serious ghost artefacts will appear in the moving regions of the generated HDR image. In this paper, we propose a novel ghost-free exposure fusion method based on photometric relation to detect moving regions in the dynamic scene. Our method makes full use of the linear relation between brightness and exposure time, instead of only using photometric order relation. Our method is easy to implement without camera response function estimate and complicated tone mapping operations. This method detects ghost regions better than the previous photometric order relation based method and generate the fused image with less artefacts. Many experimental results with different dynamic scenes show the effectiveness of our method.