Invention Description
Detecting bright point scatterers is important to quality assessments of sonar, radar and medical imaging systems, particularly as it pertains to resolution characterization. Prominent, or coherent, scatterers are typically detected using thresholding techniques in addition to statistical measures in the detection processing chain. These methods, however, perform poorly in detecting point-like scatterers when there are high levels of speckle background. Further, they may distort the structure of the scatterer when visualized.
Researchers at Arizona State University have developed a novel image processing technique derived from analyzing ideal point scatterer responses to enhance bright features in synthetic aperture sonar (SAS) and radar (SAR) images, especially in the presence of high speckle noise. This bright feature transform (BFT) is an analytic, computationally efficient technique which employs a sine-based tone mapping function via element-wise multiplication, to highlight scatterers without distorting their structures. It uses a novel trigonometric difference approach to suppress speckle noise and preserve scatterer structure and width during tone mapping. Evaluations on simulated and real datasets demonstrate superior results compared to traditional thresholding and advanced deep learning methods, delivering higher image quality, detection accuracy, and faster processing.
The method provides a fast, threshold-free technique to detect prominent scatters in SAS and SAR imagery with high noise while preserving the overall shape of the prominent scatterer for more accurate downstream analyses.
Potential Applications
- Marine and underwater exploration using synthetic aperture sonar
- Remote sensing and surveillance with synthetic aperture radar
- Medical ultrasound imaging
- Environmental monitoring and mapping
- Defense and security applications requiring precise target detection
- Preprocessing for machine learning and image analysis workflows
Benefits and Advantages
- Minimal parameters and computationally efficient
- Requires no threshold tuning
- Preserves the shape and edges of bright scatterers without distortion
- Enhanced noise mitigation through trigonometric-based functions
- Suppresses speckle and background noise
- Applicable to both raw amplitude and log-compressed imagery
- Robust detection in high noise environments
- Enhances resolution and sharpness for downstream image analysis
- Outperforms traditional thresholding and some deep learning approaches
- Faster than standard denoising techniques with comparable or better results
- Versatile use for visual analysis and preprocessing for downstream tasks
For more information about this opportunity, please see