Related Experiment Video
Updated: Mar 19, 2026
10:37
A Multimodal Imaging Framework to Advance Phenotyping of Living Label-free Breast Cancer Cells
Published on: August 22, 2025
1.4K
Blind Source Parameters for Performance Evaluation of Despeckling Filters
Nagashettappa Biradar1, M L Dewal2, ManojKumar Rohit3
1Bheemanna Khandre Institute of Technology, Bhalki 58532, India.
International Journal of Biomedical Imaging
|June 15, 2016
Summary
Evaluating echocardiographic image filters is challenging due to inherent speckle noise. Blind assessment metrics like the speckle suppression and mean preservation index (SMPI) offer effective evaluation without noise-free references.
Area of Science:
- Medical Imaging
- Signal Processing
- Ultrasound Technology
Background:
- Speckle noise is an inherent artifact in transthoracic echocardiographic images.
- Traditional image quality metrics (PSNR, MSE, SSIM) are inadequate for evaluating despeckling filters on echocardiograms due to the absence of noise-free references.
- Blind assessment metrics are crucial for evaluating filter performance on echocardiographic data.
Purpose of the Study:
- To comprehensively analyze and evaluate eleven despeckling filters for echocardiographic images.
- To compare filter performance using both traditional and blind assessment metrics, including clinical validation.
- To identify optimal filters for reducing speckle noise while preserving essential image features.
Main Methods:
- Eleven despeckling filters were evaluated on echocardiographic images.
- Performance was assessed using traditional metrics (PSNR, MSE, SSIM) and blind metrics (speckle suppression index, SMPI, beta metric).
- Clinical validation was performed to assess the practical utility of the filters.
Main Results:
- Logarithmic neighborhood shrinkage (NeighShrink) with SURE effectively suppressed speckle noise.
- The SMPI demonstrated superior performance, being three times more effective than the wavelet-based generalized likelihood estimation approach.
- Filters like nonlocal mean, Bayesian estimation, hybrid median, and probabilistic patch-based methods were found to be acceptable for echocardiographic images.
- Median, anisotropic diffusion, fuzzy, and Ripplet nonlinear approximation filters showed limited applicability.
Conclusions:
- Blind assessment metrics overcome the need for noise-free reference images in echocardiography.
- Specific filters, including nonlocal mean and Bayesian methods, are recommended for effective speckle reduction in echocardiographic imaging.
- Certain filters have limited clinical utility for despeckling echocardiograms, necessitating careful selection based on performance metrics.
