A locally adaptive approach for interferometric phase noise reduction

This paper presents a new filtering technique for SAR interferometric phase images. Due to the presence of noise in the data, a filtering step must be performed before unwrapping the interferogram in order to obtain a more accurate evaluation of the true phase values and, as a consequence, a better topographic model. The goal of this technique is to reduce the phase noise that has an additive gaussian model, preserving the local contrast and the shape of the fringes. Noise reduction should be realized involving an adaptive approach that takes into account the aspect and the local statistics of the interferometrical image. The authors propose an edge-preserving method that exploits the local content of the phase image in order to determine for each pixel the best matching shape of the filtering mask, searching in a set of different neighbourhood systems. To test whether a proposed mask is efficient or not, the local variance is computed for each neighbourhood a proposed mask shape at a certain pixel position is accepted only if its variance value is lower than those of alternative shapes. Once the best mask has been chosen, a nonlinear adaptive filtering function can be applied. The effectiveness of this technique can be seen in a considerable reduction of the interferometric noise, while preserving the integrity of phase gradients, i.e., the contrast and the morphology of fringes.