A novel post-nonlinear ICA-based reflectance model for 3D surface reconstruction

We propose a novel reflectance model for photometric stereo. It consists of diffuse components and specular components. Unlike past methods, ours does not need to separate the two components from the nonlinear reflection model. We use an unsupervised learning adaptation algorithm to estimate the reflectance model based on image intensities. First, the technique of the post-nonlinear independent components analysis (ICA) model is used to obtain the surface normal on each point of an image. Then, the 3D surface model can be reconstructed based on the estimated surface normal on each point of the image by using the method of enforcing integrability. We test our algorithm on synthetically generated images for the reconstruction of the surface of objects and on a number of real images from the Vale Face Database B. The results clearly indicate the superiority of the proposed nonlinear reflectance model over the Georghiades approach and the Hayakawa approach.