Diffusion equation based parameterization of light field and computational imaging model

Abstract

The parameterization of light field is fundamental for the light field data acquisition and the computational imaging model. In this paper, we proposed the diffusion equation based parameterization of light field and established the computational imaging model via the proposed parameterized light field. Since the light propagation in the image space can be regarded as a diffusion process, the light field in the image space can be represented based on the diffusion equation in the lens imaging. The computational imaging model via the proposed light field parameterization consists of the forward model and the inverse problem. The forward model is the generation of the focal stack from the depth map and the all-in-focus (AIF) image. The inverse problem is to reconstruct the depth map and the AIF image from the focal stack. We clarify that the focal stack can be regarded as the scale spaces of the local region images. As a result, the depth reconstruction and the all-in-focus imaging from focal stack can be solved by detecting the extremum of the feature point density. In this paper, the simulated focal stack and the real captured focal stack data are used to verify the reconstruction method, and the high-precision scene depth map and the all-in-focus image are reconstructed.