Accurate and efficient extraction of fringe orientation from the poor-quality ESPI fringe pattern with a convolutional neural network

The fringe orientation is an important feature of the electronic speckle interferometry (ESPI) fringe pattern. Accurate and efficient calculation of the fringe orientation is very important for subsequent electronic speckle processing such as skeleton extraction and image filtering. To accurately and efficiently estimate fringe orientation, we propose an effective method based on a convolutional neural network. In the proposed method, the network needs clean-noisy image pairs to train and noisy images with theoretical value to test. The aligned noise-free ESPI fringe pattern orientation fields are fairly good estimations for the corresponding noise ones. After the model training is done, the other multiple ESPI fringe patterns are fed to the trained network simultaneously; the corresponding orientation results can be obtained accurately and efficiently. The advantage of using this method to extract the orientation is that the fringe orientation information can be extracted accurately and efficiently without complicated parameter adjustment. We evaluate the performance of our method via applying our method to the computer-simulated and experimentally acquired ESPI fringe patterns and comparing the results with those of three extensively used methods.