Stochastic Systems Group
Home Research Group Members Programs  
Demos Calendar Publications Mission Statement Alumni

SSG Seminar Abstract


Factor Graph and Optimization Method for a New Phase-Unwrapping Algorithm

Jonathan Yedidia
Mitsubishi Electric Research Laboratory


I will describe a new "Iterated Conditional Modes / Minimum Cost Flow" (ICM-MCF) algorithm for phase-unwrapping in the presence of noise. The ICM-MCF algorithm finds a locally optimal configuration of unwrapped phases under a well-characterized factor graph model of the terrain and noise.

This algorithm has been integrated into Mitsubishi Electric's airborne interferometric Synthetic Aperture Radar (inSAR) system, and has considerably improved the performance of the overall inSAR system, as shown by recent field tests, which recovered a digital surface model with an error of less than 50 cm in height at a mesh resolution of 50 cm X 50 cm.



Problems with this site should be emailed to jonesb@mit.edu