Related Experiment Videos
Segmentation of speckle images based on level-crossing statistics
1Department of Electrical Engineering, Rochester Center for Biomedical Ultrasound, University of Rochester, New York 14627.
Summary
This study introduces a novel method for segmenting images affected by speckle noise. By analyzing clipped speckle images using level-crossing statistics, it accurately identifies regions for mean value estimation in coherent imaging systems.
Area of Science:
- Image processing
- Coherent imaging systems
- Stochastic processes
Background:
- Speckle is a common artifact in coherent imaging, arising from the stochastic nature of wave interference.
- Accurate estimation of image intensity or envelope mean values requires ergodic regions.
- Existing methods struggle with segmenting these regions from single speckle realizations.
Purpose of the Study:
- To develop a new image segmentation method for speckle noise.
- To accurately identify regions for mean value estimation in single-realization speckle images.
- To address the challenge of region identification in stochastic imaging processes.
Main Methods:
- Clipping the speckle image at a constant threshold to create a bilevel image.
- Analyzing the level-crossing statistics of the resulting bilevel image.
- Employing morphological transformations (opening and closing) to measure feature sizes.
Main Results:
- A decision rule for region segmentation based on speckle fades and excursions was derived.
- The method was successfully applied to computer-generated speckle images.
- The approach demonstrated effective identification of distinct image regions.
Conclusions:
- The proposed level-crossing statistics method offers a robust approach to speckle image segmentation.
- This technique can improve mean value estimation in applications with speckle phenomena.
- Potential applications include laser, ultrasound, and radar imaging.