Spatial vector filtering to reduce noise in interferometric phase images

Spatial filtering to reduce noise is an important process to improve phase measurements in interferometric imaging. Currently these techniques are predominately based on filtering the interferometric phase. However the amplitude, which is ignored or only used to a minimal extent in conventional filtering, holds useful information about the nature of the phase measurement. When combining multilook images, adding the full complex vectors produces a least mean square estimate of the interferometric phase. Utilising this principle, a new spatial filtering technique is introduced. In this technique, a compromise is suggested where the square-root of the interferometric amplitude is taken prior to summing the spatially weighted vectors. To deduce the weights for the filtering function it was necessary to utilise a Monte-Carlo approach, which calculates the filter weights via an iterative minimisation procedure. The main application for this filter is to improve the quality of displacement maps obtained by a ground based interferometric radar system that monitors rock slope movement in a mining environment. For this application high SNR is expected, however, due to speckle, large errors can still be present in the image. The filter is designed to reduce these errors. Using simulated data and C-Band ERS images, the new spatial vector filter is shown to produce improved phase measurements compared with conventional filtering techniques