Wrapped polygon methods, faster copy and new skeletonization algorithm

[valillon]

Hi @kylemcdonald ,

Recently using your addon, I come across with some needs and finally ended up implementing them.
Major changes in this request:

• Added wrappers for fillConvexPoly() and pointPolygonTest().
• Faster cv::Mat copy in CLD method by using opencv pointers.
• New skeletonization algorithm based on the Zhang-Suen method. It's slower but more robust. Actually I found some wrong behavior in the current ofxCv::thin(). I leave it there in the commit since it is faster and can works fine for simple cases.

[roymacdonald]

Hi @valillon, I found this same odd behavior some time ago, made some fixes but ended up agreeing with @kylemcdonald that it was not a good fix. look here https://github.com/kylemcdonald/ofxCv/issues/183 and it's related PR https://github.com/kylemcdonald/ofxCv/pull/184 What's currently implemented is not doing the skeletonization properly It is not doing all the iterations it needs, so that's probably the reason for such a big difference in the processing times. I'll try out your pr and let you know. cheers

[valillon]

Hi Roy!,

Actually Kyle already mentioned the same Zhang-Suen algorithm in this comment. Thanks for testing.

[roymacdonald]

Hey @valillon much better then! I will test soon and let you know. best!

[valillon]

Hi again! I just did some simple optimizations to the Nash's skeletonize code in my local branch here. It depends much on the complexity of the binary image, but at least for the images in example-coherent-lines it performs about x2 faster.

[kylemcdonald]

thank you! this generally looks good. thanks for checking @roymacdonald since i know you've thought about this before.