A new algorithm for adaptive non-linear digital filtering based on first order local statistics is proposed in this paper. A pixel value is replaced with either of the percentile thresholds of the local histogram only if it belongs to the corresponding percentile tail. Percentiles are adaptively defined as functions of the local variation coefficient in such a way to introduce a selective smoothing cffect. The filter is tunable for more or less noisy images and maintains the local mean same as a linear filter. Performances have been evaluated on polarimetric SAR L, P, C, band data of Montespertoli test site collected during the NASA Multisensor Aircraft Campaign (MAC 91) worked out in June-July 1991. Photointerpretation results are improved since speckle is reduced while contours and structures are preserved.
A New Adaptive Digital Filter for SAR Images: Test of Performance for Land and Crop Classification on Montespertoli Area
L Alparone;S Baronti;
1992
Abstract
A new algorithm for adaptive non-linear digital filtering based on first order local statistics is proposed in this paper. A pixel value is replaced with either of the percentile thresholds of the local histogram only if it belongs to the corresponding percentile tail. Percentiles are adaptively defined as functions of the local variation coefficient in such a way to introduce a selective smoothing cffect. The filter is tunable for more or less noisy images and maintains the local mean same as a linear filter. Performances have been evaluated on polarimetric SAR L, P, C, band data of Montespertoli test site collected during the NASA Multisensor Aircraft Campaign (MAC 91) worked out in June-July 1991. Photointerpretation results are improved since speckle is reduced while contours and structures are preserved.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
