Adaptive pixel classifier for binary structured light: A probabilistic kernel approach

The paper proposes an adaptive classification mechanism designed for structured light system to improve quality of reconstructed models. We observed that the conventional albedo-based thresholding fails when the lighting condition is not carefully considered. To address this problem, an adaptive model is proposed. The core idea is to adjust decision boundary during extraction of sequence of binary-coded light patterns by taking the change of lighting condition into account. Base on this idea, a probabilistic kernel-based online learning procedure has been designed and applied to a structured light system. The experimental results show that the proposed method yields more reliable pixel classification as well as increased accuracy of the 3D scanner. It should be noted that the proposed method does not require any modification on conventional Gray-coded patterns.