Multi-exposure image fusion method using anisotropic diffusion

Abstract

Image fusion involves selective combination of two or more complementary images from multiple sources/sensors in order to achieve a single image which carries relevant information regarding the targeted scene/region. The paper focuses upon multi-exposure fusion methodology—herein a targeted scene is captured by stationary camera with different exposure times. This is followed by fusion of these generated multi-exposure images using nonlinear Anisotropic Diffusion (AD). The significance of AD filtering lies in its capability of image decomposition into homogeneous and non-homogeneous regions. This very attribute supports pre-filtering and preservation of contrast and details during the fusion. The fusion results are then validated using state-of-the-art fusion metrics to justify the performance and benchmarking.