Batch denoising of ESPI fringe patterns based on convolutional neural network.

Area of Science:

• Optics and Photonics
• Image Processing
• Artificial Intelligence

Background:

• Electronic Speckle Pattern Interferometry (ESPI) is a powerful optical technique for measuring surface deformation.
• Denoising ESPI fringe patterns is crucial for accurate analysis but challenging due to noise and low contrast.
• Existing denoising methods may struggle with large datasets or low-quality images.

Purpose of the Study:

• To develop an automated and efficient method for denoising ESPI fringe patterns.
• To leverage Convolutional Neural Networks (CNNs) for batch processing of ESPI images.
• To create a robust training dataset using a novel computer-simulated ESPI fringe pattern generation method.

Main Methods:

• A CNN model was trained on a custom dataset of noisy and noise-free simulated ESPI fringe patterns.
• The trained CNN was applied to denoise multi-frame ESPI fringe patterns in batches.
• Performance was evaluated using both computer-simulated and experimentally acquired ESPI fringe patterns.

Main Results:

• The proposed CNN method effectively denoised ESPI fringe patterns, achieving desired results even with low-quality images (high noise, low contrast).
• The method demonstrated simultaneous denoising of multi-frame ESPI patterns.
• The trained network successfully processed both simulated and experimental ESPI fringe patterns.

Conclusions:

• The CNN-based batch denoising method offers a significant advancement for ESPI applications, particularly for large-scale data processing.
• The approach is robust and effective even for challenging image conditions.
• The use of simulated data for training enables broad applicability to real-world ESPI data.