Related Experiment Videos
BM3D adaptive TV filtering-based convolutional neural network for ESPI image denoising.
Applied Optics
|February 24, 2022
Summary
This study introduces an improved denoising convolutional neural network (CNN) for electronic speckle pattern interferometry (ESPI) images. The method effectively reduces speckle noise while preserving crucial image details and texture.
Area of Science:
- Image Processing
- Computational Imaging
- Optical Metrology
Background:
- Speckle noise in electronic speckle pattern interferometry (ESPI) degrades image quality and hinders analysis.
- Effective noise reduction is crucial for accurate interpretation of ESPI data.
Purpose of the Study:
- To propose an improved denoising convolutional neural network (CNN) for reducing speckle noise in real ESPI images.
- To preserve essential image details, texture, and edge information during the denoising process.
Main Methods:
- Development of an improved denoising CNN incorporating block matching 3D-based adaptive total variation (TV) denoising.
- Implementation of a two-channel model to enhance noise reduction performance.
- Comparative analysis against conventional and deep-learning denoising algorithms.
Main Results:
- The proposed method significantly reduces speckle noise in ESPI images.
- Preservation of image accuracy, integrity, and stability was demonstrated.
- Details, texture, and edge information of the stripe patterns were effectively maintained.
Conclusions:
- The improved denoising CNN offers a robust solution for enhancing ESPI image quality.
- The method balances effective noise suppression with the preservation of critical image features.
- This approach is beneficial for subsequent quantitative analysis of ESPI measurements.
