Journal of the European Optical Society-Rapid Publications, Volume. 19, Issue 1, 2023015(2023)
Convolutional neural network optimisation to enhance ESPI fringe visibility
Fig. 2. Example U-NET network with an input image of 256 × 256 and 1 channel (B/N) and 32 × 32 resolution after the full encoder path. The decoder path reverses the encoding operations and uses inputs from the corresponding encoder layer (skip connections) to end with a cleaned output image.
Fig. 3. Generated image samples. (A) Computed ESPI image using only the n = 3, 8 and 14 first Zernike polynomials to simulate the complexity of specimen displacement. (B) Clean image or ground true
Fig. 4. U-NET network finally used concatenating blocks consisting of two 5 × 5 convolutions (each one followed by a ReLU activation unit) and a 2 × 2 maxpooling operation with stride = 2 for downsampling along the encoding path, reversing the operations along the decoder path using blocks composed of upsampling operations followed by a 2 × 2 convolution (up-conv), concatenated with the corresponding output of the encoding part and followed by two 5 × 5 convolutions (each one followed by a ReLU activation unit).
Fig. 5. Samples of cleaned images using the selected hyperparameters. (A) Input image. (B) Expected output (Ground true). (C) Processed image by the U-NET. The SSMI index is computed using expected and processed image (columns B and C).
|
Get Citation
Copy Citation Text
José Manuel Crespo, Vicente Moreno. Convolutional neural network optimisation to enhance ESPI fringe visibility[J]. Journal of the European Optical Society-Rapid Publications, 2023, 19(1): 2023015
Category: Research Articles
Received: Jan. 31, 2023
Accepted: Mar. 23, 2023
Published Online: Aug. 31, 2023
The Author Email: Crespo José Manuel (josemanuel.crespo.continas@rai.usc.es), Moreno Vicente (josemanuel.crespo.continas@rai.usc.es)