Abstract: The phase information of magnetic resonance (MR) phase image can be used in many MR imaging techniques, but
phase wrapping of the images often results in inaccurate phase information and phase unwrapping is essential for MR
imaging techniques. In this paper we analyze the causes of errors in phase unwrapping with the commonly used Goldstein
Brunch-cut algorithm and propose an improved algorithm. During the unwrapping process, masking, filtering, dipoleremover
preprocessor, and the Prim algorithm of the minimum spanning tree were introduced to optimize the residues
essential for the Goldstein Brunch-cut algorithm. Experimental results showed that the residues, branch-cuts and continuous
unwrapped phase surface were efficiently reduced and the quality of MR phase images was obviously improved with the
proposed method.