**6. Conclusions**

Computational power has been increased in the last years. This increased processing power allows to develop EA as a practical tool with application in patter recognition, computational intelligence, image processing, automatization, and others. EA's, fuzzy logic and neural networks are called soft computing, because they can deal with problems where there are not a good knowledge of the problem, information is incomplete or inconsistent, or are large amount of noise.

In this chapter is shown only a single application of EA on fringe pattern demodulation. There still a lot of variations that can be explored to improve the performance of actual algorithms.

GA based methods have two advantages over regularized phase tracker: they can work on low resolution images and they can follow changes in concavity. These advantages are the consequence of taking upper grade terms in the interpolated function.

The technique showed in (Toledo, Cuevas 2009), called FPIW, is based in two suppositions: it is not necessary to know the phase on the neighborhood to estimate the phase in a given window, and the estimated phase in a window differ only by its concavity sign and a DC bias, from the real phase in the region framed by the window. As a consequence, the overlapped similarity criterion used in the WFPD (Cuevas et al, 2003) method can be eliminated from the fitness function in the FPIW method. In exchange, FPIW works near Nyquist, but on sub-Nyquist, WFPD is better.

The phase in a given window is estimated without known nothing about the phase in other windows. It is possible to demodulate simultaneously all windows, that is, FPIW method described has implicit parallelism. WFPD demodulate the windows sequentially.
