Three-dimensional Integral Imaging Visualization in Scattering Medium with Baysian Estimation

Abstract

We propose an image descattering method for improved image visualization in scattering medium such as heavy turbid water. The descattering method is applied to three-dimensional (3D) integral imaging (InIm). The 3D InIm pick-up process captures multiple two-dimensional images of a scene each having a different perspective of the scene, known as elemental images (EIs), and combines this information to generate a 3D reconstruction of the scene. A scattering mitigation process is applied to each EI prior to 3D reconstruction to reduce the effect of scattering. In the scattering mitigation process, we assume that the intensity of the scattering medium containing object information is a Gaussian distribution and the distribution of the scattering medium is known a priori and also follows a Gaussian distribution for the whole captured image. By computing maximum a posteriori estimates of the mean and variance of the turbid media containing object information, a Bayesian scattering mitigation process is implemented. After the scattering mitigation process, the 3D reconstruction image is calculated computationally using back propagation method. We compare the reconstructed image quality of an existing method and our proposed method using structural similarity index (SSIM), and show our proposed method achieve higher SSIM.