Fast Subpixel Accurate Reconstruction Using Color Structured Light

Extracting range data through active optical triangulation requires robust stripe edge detection of the emitted pattern because of undesired shape and shading variations in the scanned object. We propose several extensions to a structured light system originally proposed by Zhang et al. [1]. Assuming stripes of a certain width, our edge detection criterion is achieved simply by restricting the number of consecutive edge pixels. Subpixel accuracy from one single input image can be obtained by approximating the gradient of the square local contrast with a piecewise linear function. Finally, we present a noise reduction technique for the range map through meshing followed by a parameterized face orientation culling. Experimental results have demonstrated major improvements in terms of robustness against acquisition noise, shading variations and complex shapes.