An active stereo method for three-dimensional shape measurement is proposed. The proposed method uses epipolar constraint and binary coding structured light to perform dense and accurate stereo matching. The correspondence in horizontal directions is determined by epipolar constraint, and the binary code determines the correspondence in vertical directions. A high redundancy skeleton coding method is proposed to improve the decoding accuracy significantly. And the matching algorithm is achieved sub-pixel accuracy. The point cloud is calculated by triangulation. Spatial re-sampling based on the Delaunay triangulation and inverse distance weighting interpolation is implemented to get a uniform data density for a comfortable visualization. Compared with phase-shift methods, the proposed one is more robust. It preforms reliably even in noisy conditions. The experiment results are presented to demonstrate the effectiveness of the proposed method. The resolution is analyzed and verified by experiment. It achieves 1/1200 by two 2-megapixel cameras and can improve by increasing the image resolution. The robustness is also verified. The experiment demonstrated that our method has great immunity of three main type of noise in the phase-shift method.