Structured-light 3D sensors are widely used in reverse engineering, medical imaging and digital archive. Especially, structured-light methods adopting binary patterns are good for reconstructing accurate 3D structures of static objects so that many scanners in the market are using binary patterns. However, when image noise and inter-reflection on the surface occur, it can be failed to estimate correct binary codes only which can be utilized to calculate precise 3D structures. In this paper, we propose a novel filtering algorithm that can eliminate miss-estimated binary codes caused by noise and inter-reflection. The experimental results show the effectiveness of the proposed algorithm.