Land-cover separability studies of filtered ERS-1 SAR images in the tropics

The study compares various SAR filtering techniques in the spatial domain to investigate the separability of different land cover categories. Eight commonly applied different filters are implemented and tested at a standard ERS-1 SAR image taken over Thailand. Homogeneous areas of seven representative land-cover categories are selected, namely paddy, sugarcane, shrubs, bushfallow, urban areas, and waterbodies. Mean backscattering coefficients are derived before and after speckle filtering. The evaluation of the filters is primarily based on the improvement of class separability after filtering. In addition, improvement of signal to noise ratio after filtering and visual interpretation of the filtered versus unfiltered image are investigated. The study further compares a single date, unfiltered with an optimally filtered ERS-1 SAR image in respect to land-cover separability based on digital classification and visual analysis.<
>