2020/10/25 by Malsha V. Perera, Perera, Malsha V., Ashwin De Silva +1 · 1 citation
Computer Science · Engineering · #FOS: Computer and information sciences #FOS: Electrical engineering #Machine Learning (cs.LG) #Optical measurement and interference techniques #Signal Processing (eess.SP) #Structural Health Monitoring Techniques #Ultrasonics and Acoustic Wave Propagation #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2010.13268
openalex publication_date 2020/10/25 · openalex created_date 2022/07/25 · openalex updated_date 2026/07/28
Phase unwrapping is a classical ill-posed problem which aims to recover the\ntrue phase from wrapped phase. In this paper, we introduce a novel\nConvolutional Neural Network (CNN) that incorporates a Spatial Quad-Directional\nLong Short Term Memory (SQD-LSTM) for phase unwrapping, by formulating it as a\nregression problem. Incorporating SQD-LSTM can circumvent the typical CNNs'\ninherent difficulty of learning global spatial dependencies which are vital\nwhen recovering the true phase. Furthermore, we employ a problem specific\ncomposite loss function to train this network. The proposed network is found to\nbe performing better than the existing methods under severe noise conditions\n(Normalized Root Mean Square Error of 1.3 % at SNR = 0 dB) while spending a\nsignificantly less computational time (0.054 s). The network also does not\nrequire a large scale dataset during training, thus making it ideal for\napplications with limited data that require fast and accurate phase unwrapping.\n