A Non-Local despeckling approach using pixel-similarity thresholding

• Published in: IET Conference Proceedings p. B0670
• Main Authors: Li Guangting, Dang Hongxing, Tan, Xiaomin, He Yapeng
• Format: Conference Proceeding
• Language: English
• Published: Stevenage, UK IET 2013; The Institution of Engineering & Technology
• Subjects: Optical, image and video signal processing; Other topics in statistics; Radar equipment, systems and applications; pixel relativity model; SAR image look number; pixel similarity thresholding; quantification; nonlocal despeckling approach; reflectivity estimation; ratio distance; visual quality; reflectivity; joint probability density function; neighboring reflectivity ratio; synthetic aperture radar; image segmentation; minimum error probability criterion; radar imaging; similar pixel selection; searching window; error statistics
• ISBN: 1849196036; 9781849196031
• DOI: 10.1049/cp.2013.0210

We propose a Non-Local despeckling approach by using the new pixel-relativity measure, based on ratio distance, and thresholding the pixel-similarity. Firstly, the pixel-relativity model based on ratio distance is obtained by transforming the joint probability density function of two pixels having the same reflectivity. Then a table of pixel-similarity threshold, as a function of the SAR image look number and neighboring reflectivity ratio, is trained according to the minimum error probability criterion. Finally, the pixel-similarity threshold is used to select similar pixels from the searching window for real reflectivity estimation. The proposed approach is verified by the comparison with PPB and LHRS-PRM filters using synthetic and real SAR images. The visual quality and the quantification comparison show that our method is excellent in restoring not only the uniform area but also the borders and texture area. (6 pages)