Adding Bucket Distance Hashing to OpenCV

[tomoaki0705]

Let me start the discussion here.

I'm planning to add new method for feature matching, BDH (Bucket Distance Hashing) which which was announced in ICCV 2013

https://github.com/opu-imp/BDH paper: Masakazu Iwamura, Tomokazu Sato and Koichi Kise "What Is the Most Efficient Way to Select Nearest Neighbor Candidates for Fast Approximate Nearest Neighbor Search" Ref: http://www.m.cs.osakafu-u.ac.jp/cgi-bin/publication_public.cgi?func=show_detail&id=1357

It has better performance compared to other papers.

Now since it's a new feature, I'd like to break down the task before starting. As far as I know, I should do following points.

• [ ] Porting
  • [x] Where to port The contribution guideline says I should put this in to opencv_contrib, is this correct ? The method is very similar to FLANN, so I'm just wondering if we could merge it together. (is that too aggresive modify ?)
  • [x] Which branch to commit I'm planning to push this feature in to master branch (4.x series). Is this OK ?
• [ ] Test Each method should have a regression test, so I need to add a gtest based test.
• [ ] Documents Explanation of API. Should be generated by doxygen
• [x] Patent The method should be patent free

Is there any other task I should do for PR ? Best.

[alalek]

• It is better to keep FLANN code untouched (it is 3rdparty code).
• opencv_contrib is the right place for new "standalone" algorithms. If this algorithm should be in other OpenCV places then "it depends".
• You can use master branch. Use "3.4" branch if you are sure that "side effect" (compatibility issues) would be minimal and you don't use C++11.
• Please add at least one simple test.