A twofold symmetry based approach for the feature detection of pseudorandom color pattern

Abstract

Spatially encoded structured light technique is an important means for the dynamic 3D reconstruction. This paper first present a novel designed pattern, which adopts four color elements and get a 2×2 unique window property. To detect the pattern features, i.e. the grid points, a twofold symmetry based feature detection method is investigated. The value of each pixel is firstly computed in R, G, B channels respectively. And the largest one is select to represent the pixel. By selecting the local maximum points as feature candidate, a measurement of twofold rotation symmetry is proposed to localize the grid point precisely. With comparison with some classical detectors and experiment with real human face, the robustness and accuracy of the proposed detector is demonstrated.