Hybrid method for multi-exposure image fusion based on weighted mean and sparse representation

We propose a hybrid method for multi-exposure image fusion in this paper. The fusion blends some images capturing the same scene with different exposure times and produces a high quality image. Based on the pixel-wise weighted mean, many methods have been actively proposed, but their resultant images have blurred edges and textures because of the mean procedure. To overcome the disadvantages, the proposed method separately fuses the means and details of input images. The details are fused based on sparse representation, and the results keep their sharpness. Consequently, the resultant fused images are fine with sharp edges and textures. Through simulations, we show that the proposed method outperforms previous methods objectively and perceptually.