Absolute Phase Recovery of Single Frame Composite Image Based on Convolutional Neural Network

In this paper, a convolutional neural network is proposed to obtain high quality absolute phase from single frame composite images. The composite image used in the proposed method is the fringe image embedded with speckle. The convolutional neural network consists of two sub-networks, which use the fringe mode component and the speckle mode component in the composite image to solve and unfold the wrapping phase. In the process of phase unwrapping, the proposed method uses the pre-photographed composite image and its fringe order as auxiliary information to ensure the accuracy of phase unwrapping. Experimental results show that the proposed method can minimize the number of projected images by using single-frame composite images and obtain high precision absolute phase, which provides a feasible solution for 3D measurement in high precision dynamic scenes.