Phase unwrapping is widely applied in a variety of fields including optical and microwave interferometry, adaptive optics, synthetic aperture radar and image processing. A phase unwrapping method based on segmenting regions containing phase jumps is proposed. The more the neighbors around a pixel have phase values near π and -π, and the smaller the number difference between the neighbors near π and -π is, the more possibly the pixel locates near a phase jump line. The pixels with high possibility form the phase jump regions, which can be eliminated by compensating the pixels with a multiply of 2π to minimize the phase difference between them and their neighbors. After removing phase jumps, ordinary filtering operations can be utilized to postprocess the noisy unwrapped phase map. Computer simulation and experimental results make it clear that the method can effectively remove noise in phase map and unwrap the wrapped phase.