We describe what is to our knowledge a novel technique for phase unwrapping. Several algorithms based on unwrapping the most-reliable pixels first have been proposed. These were restricted to continuous paths and were subject to difficulties in defining a starting pixel. The technique described here uses a different type of reliability function and does not follow a continuous path to perform the unwrapping operation. The technique is explained in detail and illustrated with a number of examples.

We describe what is to our knowledge a novel approach to phase unwrapping. Using the principle of unwrapping following areas with similar phase values (homogenous areas), the algorithm reacts satisfactorily to random noise and breaks in the wrap distributions. Execution times for a 512 × 512 pixel phase distribution are in the order of a half second on a desktop computer. The precise value depends upon the particular image under analysis. Two inherent parameters allow tuning of the algorithm to images of different quality and nature.

In phase unwrapping residues are points of locally inconsistent phase that occur within a wrapped-phase map, which are usually regarded as being problematic for phase-unwrapping algorithms. Real phase maps typically contain a number of residues that are approximately proportional to the subsequent difficulty in unwrapping the phase distribution. This paper suggests the radical use of the discrete Fourier transform to actually increase the number of residues in 2D phase-wrapped images that contain discontinuities. Many of the additional residues that are artificially generated by this method are located on these discontinuities. For example, in fringe projection systems, such phase discontinuities may come from physical discontinuity between different parts of the object, or by shadows cast by the object. The suggested technique can improve the performance of path independent phase-unwrapping algorithms because these extra residues simplify the process of setting the branch cuts in the wrapped image based on the distance to the nearest residue. The generated residues can also be used to construct more reliable quality maps and masks. The paper includes an initial analysis upon simulated phase maps and goes on to verify the results on a real experimental wrapped-phase distribution.

A new technique for fringe pattern demodulation using a linear digital phase locked loop (DPLL) is presented. The phase components of a fringe pattern are calculated using a way similar to the Fourier fringe analysis technique and a wrapped phase map is produced. The linear DPLL demodulates, tracks and unwraps the wrapped phase map simultaneously. Consequently, the proposed algorithm has good noise immunity. The proposed algorithm is tested using real fringe patterns and its suitability for demodulating real fringe patterns is confirmed. Also, the proposed algorithm is compared with the Fourier fringe analysis technique. The implementation of the linear DPLL in real time on a commercial parallel image processing system is described. Fringe patterns are grabbed, processed and the resultant phase maps are displayed concurrently. The parallel system is capable of analysing 25 images per second. There is a two-image delay between grabbed fringe patterns and displayed phase maps.