Patch-based Laplacian filter based transmission estimation for single image dehazing

Transmission estimation is the primary task for single image haze removal. In this paper, we proposed a more accurate transmission estimation method for image haze removal based on any transmission extraction prior (eg., dark channel prior, color line prior). Firstly, we generate raw transmission in Gaussian pyramid using any transmission extraction prior. Secondly, a patch-based Laplacian pyramid is built from the nonlinear remapped function point by point. Thirdly, we obtain the global Laplacian map with patch-based remapped image and use them to guide the transmission refinement. Finally, the dehazing result is restored from the refined transmission and atmospheric scattering model. Experiments show that the proposed method achieves higher quality of halo-free and detail restoration performance than the existing dehazing methods.