Deep Learning-Based Phase Unwrapping Method
Dongxu Li, Xianming Xie
Abstract
A phase unwrapping method based on spatial and channel attention network is proposed to retrieve true phases from interferograms with various levels of noise. First, we propose a network that is suitable for unwrapping wrapped phase images. This network utilizes Deeplabv3+ as the backbone, adopts a serial-parallel atrous spatial pyramid pooling module, implements multi-scale skip connections between the encoder-decoder models, and fuses a convolutional block attention module. Second, datasets with different noise levels are used to train the network employing an existing noise level evaluation system, and the trained networks effectively handle the phase unwrapping for interferograms. Finally, the interferograms are unwrapped by the networks with the same noise level as the interferograms. The experimental results of phase unwrapping for interferograms fully verify the performance of this method.