Convex optimization for fast image dehazing

This paper presents a new convex optimization method for the image dehazing problem. It is based on reformulation of the hazed image model to solve the bilinearly coupled hazed image and light transmission distribution term as a single optimization variable. This resolves the nonconvex difficulty of the original problem and is sufficient for straightforward reconstruction of the haze free image. The Haar wavelet transform is further applied to derive a subband hazed image model for the convex optimization, which leads to great reduction of the computational workload.