Denoising by coupled partial differential equations and extracting phase by backpropagation neural networks for...

Area of Science:

• Optical Metrology
• Image Processing
• Computational Physics

Background:

• Electronic Speckle Pattern Interferometry (ESPI) is crucial for precise measurements.
• Noise and phase ambiguity in ESPI fringe patterns limit accuracy.
• Existing denoising and phase unwrapping methods have limitations.

Purpose of the Study:

• To develop an advanced denoising and enhancement model for ESPI fringe patterns.
• To introduce a novel phase unwrapping method using backpropagation neural networks (BPNN).
• To improve the accuracy and reliability of ESPI measurements.

Main Methods:

• Combining coupled nonlinear partial differential equations (PDEs) for denoising with ordinary differential equations for enhancement.
• Implementing a BPNN for phase unwrapping based on a skeleton map, avoiding traditional interpolation.
• Validating the methods on simulated and experimental ESPI fringe patterns.

Main Results:

• The coupled nonlinear PDEs model effectively removes noise from ESPI fringe patterns.
• The BPNN method provides significantly more accurate unwrapped phase values compared to traditional interpolation.
• The accuracy of the BPNN method is tunable by adjusting network parameters.

Conclusions:

• The proposed combined model offers superior denoising and enhancement for ESPI.
• BPNN-based phase unwrapping presents a more accurate and adaptable alternative to conventional methods.
• This work advances the precision of optical metrology techniques.