Quantitative Analysis of 3D Face Reconstruction using Annealing based Approach

3D shape reconstruction from 2D images is an inverse problem, and is therefore mathematically ill-posed. In most of the previous research no quantitative measures of quality of 3D reconstruction have been used. Instead visual and indirect measures such as recognition results are used as the measure of quality of 3D reconstruction. This paper presents an analysis by synthesis method for solving 3D face reconstruction problems using anatomical landmarks and intensity from 2D frontal face images. For evaluating the quality of 3D shape reconstruction two objective measures 3D shape error and 2D shape error are proposed. In order to improve the quality of 3D shape reconstruction a number of steps are proposed. Firstly, the 3D shape model is constructed by establishing a dense correspondence using rigid and non rigid surface registration. Secondly shape estimation is made robust by incorporating simulated annealing into the multidimensional amoeba optimization used for recovering shape parameters.