Depth Recovery from Defocus Images Using Total Variation

According to defocus imaging theory, the amount of blur in the defocus images varies with depth of the object in the scene. So depth in the scene can be recovered by estimating the blur with the knowledge of the lens parameters. A novel application of the total variation principle is made for the estimation of the amount of blur in the defocus images. Three main processes are included: the process of image defocusing is modeled using the model of anisotropic heat diffusion, then the depth recovery problem is translated into the minimum problem of energy functional with total variation regularization, and the gradient flow is used to seek the optimal solution. The recovery of focus image and the excess restrictions are avoided. The experiment results show that the algorithm is quite effective and depth information in edge is well retained particularly. Compared with the least square algorithm, depth recovery error with the total variation algorithm is reduced about 40%.