SARスペックルノイズ低減化に関する―手法

抄録

Synthetic Aperture Radar (SAR) is a high resolution microwave sensor, but its processed data has potential noise called“speckle noise”which degrades image quality. This noise is conspicuous when the data is processed in small number of looks.
In this paper, a new method for speckle reduction, CST filter, is proposed. Using a characteristics that a distribution of observed SAR data can be estimated by theoretical probability density function (pdf), this method determines whether the target pixel is a member of homogeneous target or not, then apply noise reduction only to the pixel which is a member of homogeneous target. This technique can be applied to the data processed in any number of looks.
The proposed method as well as five major existing methods are applied to the simulated SAR data and the JERS-1/SAR data. By comparing all filtered images, it was confirmed that the proposed method has better ability for reproduction of homogeneous targets. Although the proposed method does not use any templates to preserve structured targets, it showed almost same performance to preserve edge and point targets as the other structure preservation filters.