Multiple Intensity Differentiation-Based 3-D Surface Reconstruction With Photometric Stereo Compensation

In photometric stereo technique, multiple images are used to construct surface normal vectors and a 3-D shape is constructed thereafter. Although the lighting model is a rather simple Lambertian model in general, the quality of reconstructed 3-D shapes has been improved significantly. Recently, more general varying BRDF-based 3-D reconstruction, including specularity model is reported to be feasible. In this paper, we discuss a possible improvement in multiple intensity differentiation (MID)-based 3-D shape reconstruction using photometric stereo technique by minimizing shape distortion thus achieving better accuracy. The MID is less sensitive to albedo because of the use of reflectance ratio, but it still produces some degree of distortion due to the sharp surface angles. The main objective is to make the previous MID technique less sensitive to surface angles so that it produces a higher quality 3-D shape. Since photometric stereo does not provide depth values, we use the photometric stereo to compensate the previously obtained 3-D depth values by the MID technique. We review the image ratio method briefly followed by the discussion of surface normal calculation and experimental results.