In most phase unwrapping algorithms, the image reconstruction results are obtained by shifting the phase jumps in the wrapped phase map by 2π. The performance of such algorithms is degraded by the presence of speckle noise, residual noise, noise at the height discontinuities and holes in the wrapped phase map. Thus, a filtering operation is performed prior to the unwrapping process in order to remove the noise. However, the filtering process smears the phase jumps in the wrapped phase map and therefore causes a phase shifting error during the reconstruction process. Moreover, the noise errors, hole errors and shifting errors are accumulated path-by-path during unwrapping. Accordingly, the present study proposes a new rotation algorithm for phase unwrapping applications which resolves the noise error, the error of hole, the shifting error. Existing phase unwrapping algorithms are designed to operate only on those pixels whose phase values have no noise or holes. Or they are designed to operate the three-dimensional unwrapping paths in the row and column directions to avoid the noise or holes. By contrast, the rotation algorithm proposed in this study operates on all the pixels in the wrapped phase map, including those affected by noise or holes. As a result, the noise errors and hole errors produced in existing 2π phase shifting unwrapping algorithms are eliminated. Furthermore, since in the proposed approach, the wrapped phase map is not filtered prior to the unwrapping process, the phase shifting errors induced in existing algorithms are also eliminated. The robustness of the proposed algorithm to various noise errors, hole errors and phase shifting errors is demonstrated both numerically and experimentally.
© 2012 OSAFull Article | PDF Article