.png){kind=icon}
.png){kind=icon}
Return
Quality Enhancement of Interferometric Fringe Pattern based on Deep-Learning-Based Denoising of Combined Noise
.png){kind=icon}
PRE
AI
B
K
DOI：10.3795/KSME-A.2026.50.3.199.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Optical interferometry is a precise technique for measuring optical components by extracting phase information from fringe patterns. However, real-world measurements are affected by noise, which degrades phase accuracy. Among various noise types, the second-harmonic component is known to be dominant. This study defines combined noise as the mixture of the second-harmonic component and Gaussian noise, and proposes a deep learning-based decomposition approach for its removal. A denoising residual attention UNet (DRAUNet) model was developed to effectively decompose combined noise and enhance fringe contrast. The model demonstrated superior denoising performance through simulation experiments compared with conventional methods. Moreover, the importance of considering the second-harmonic component during denoising was verified. The proposed method was further validated using real fringe patterns obtained from silicon wafer surface measurements with a Fizeau interferometer, confirming its effectiveness and robustness.
Keywords:
Combined Noise
Deep Learning
Denoising
Fringe Pattern
Optical Interferometry
Journal
T
IF:
0.2
Papers:
87
Citations:
308
Organization
Cited Papers