:: Volume 4, Issue 2 (11-2014) ::
JGST 2014, 4(2): 25-35 Back to browse issues page
Speckle Reduction in SAR Images Based on the Soft Thresholding of Curvelet Coefficients
F. Zakeri *, M. J. Valadan Zoej, M. R. Sahebi
Abstract:   (4741 Views)
Speckle noise in radar images reduces the information that can be got from these images. This paper proposed speckle reduction in SAR images by soft thresholding of curvelet coefficients with emphasis on the preserving of edges. First, multiplicative speckle noise was transformed into an additive one by taking the logarithm of the original speckled image. Then curvelet transform was taken of logarithmically transformed image and the curvelet coefficients were thresholded by soft thresholding function. The results have been compared then to the results obtained by other widely-used adaptive filters including Frost, Gamma, Kuan, soft thresholding wavelet based filters and hard thresholding of curvelet coefficients. The results showed that the soft thresholding curvelet based filter offers better results than the mentioned filters (Mean Square Error, Normalized Mean Square Error and Mean Absolute Error indices reduced 27%,27% and 15% respectively, also Equivalent Number of Looks (ENL) increased to 27/34 and 3/61 for RADARSAT1 and ENVISAT respectively). All in all results indicated that soft thresholding of curvelet coefficients besides high reduction of speckle noise is also very powerful in keeping the edges.
Keywords: Curvelet Transform, Radar Images, Hard and Soft Thresholding, Speckle Noise
Full-Text [PDF 1973 kb]   (1109 Downloads)    
Type of Study: Research | Subject: Photo&RS


XML   Persian Abstract   Print



Volume 4, Issue 2 (11-2014) Back to browse issues page