Computer Vision Lab | Research

Separating the Reflection Components with the Use of Polarization
and Determining the Reflection Parameters

Toru Takahashi Yoichi Sato Katsushi Ikeuchi


We propose a new method for reproducing the reflection properties of the objects. First, we separate the reflection components by taking the images of a dielectric object through polarization filter under linearly polarized illumination. Second, for each separated reflection component, we determine the parameters of a reflection model (Torrance-Sparrow). In estimating the parameters, we use 3D mesh-model which is made from range images taken by a range sensor. Last, we synthesize the virtual image using the obtained reflection parameters.

Publication
1. T. Takahashi, Y. Sato and K. Ikeuchi, ``Separating the reflection components with the use of polarization and determining the reflection parameters,'' IPSJ Transactions on Computer Vision and Image Media, 2000-CVIM-124-3, Nov., 2000.


Left:Image acquisition system
Right:Separated reflection components


Results

Left:Specular parameter images
Right:Comparison between input images and synthesized images

Name, Ikeuchi Lab, University of Tokyo, 2002