may be useful: phase unwrapping

Hi,
I implemented a fast and noise-robust phase unwrapping method based on DCTs for AF's python wrapper. You can find the code and info about the corresponding [publication](http://www.mdpi.com/2313-433X/1/1/31) in the attached zip.
The DCT originates from [here](http://fourier.eng.hmc.edu/e161/lectures/dct/node2.html) and is normalized to reproduce results like [Scipy's DCT II and III](https://docs.scipy.org/doc/scipy/reference/generated/scipy.fftpack.dct.html). It interleaves the input array instead of mirror reflecting it in order to keep the FFTs the same size.
The algorithm is viable for not-too-noisy C2 smooth phase maps and can be used stand alone or as first guess for the preconditioned conjugate gradient method. When sparse support comes to AF I'll implement PCG phase unwrapping.
Use this as you see fit.

[phase_unwrapping.zip](https://github.com/arrayfire/arrayfire-python/files/224653/phase_unwrapping.zip)

Start by reviewing the attached phase_unwrapping.zip and the repository's Python wrapper to determine how, or whether, the proposed method should be exposed. Read the linked publication and DCT references, then establish the integration scope and tests needed before considering the work complete.

由索引模型根據 Issue 內容生成。