ABSTRACTFor interferometric synthetic aperture radar (InSAR) processing, the features of interferometric phase obtained in different coherence regions usually differ from each other. This is called region effect and exists in InSAR coherence map. When coherence value is used as a parameter to filter the phase noise, the result will be highly affected by this region effect. In this paper, we propose a new method of filtering InSAR phase noise using a split-window model. The idea of this method is to incorporate several filters into the model. Different filters will be used when dealing with phase noise locates in different coherence regions. The over-filtering or under-filtering caused by the coherence region effect can be eliminated in this method. As an example to demonstrate the superiority of this method, we incorporated an improved Goldstein filter and empirical mode decomposition filter into the current model. They were included to control phase noise level in the low- and high-coherence regions, respectively. The quantitative results obtained using a COnstellation of small Satellites for the Mediterranean basin Observation (COSMO-SkyMed) image pair over Kilauea volcano in Hawaiian demonstrate the advantages of the newly developed split-window model in filtering different types of noise.