Rolling ball / Sliding paraboloid algorithm
The Rolling ball / Sliding paraboloid algorithm is a well know singular background subtraction approach. It seems this would be a worthwhile addition to OpenCV. This is a fundamental and potentially very powerful approach.
In case it helps, I've already have some work in progress on this, though lack the full insight to integrate this appropriately into OpenCV. I'm happy to share this. Additional links below.
ImageJ / FIJI Java version: https://imagej.net/Rolling_Ball_Backg...
A python implemention making some use of OpenCV: https://github.com/mbalatsko/opencv-r...
Potential approximations: https://stackoverflow.com/questions/2...
isnt it more bg segmentation ? not subtraction (from many images) ?
it's a c++ library. chances, that they accept python code are low. your current implementation alse depends on various other (monster) python libs, like scikit.
https://github.com/opencv/opencv/wiki...
Thanks for the response. Without getting too much into definitions, the algorithm seems to be mainly used for foreground detection / background subtraction (https://en.wikipedia.org/wiki/Foregro...)
To clarify further: The current code I found is: 1. in Java (the original, using proprietary gfx calls), 2. A translation into Python (using some OpenCV calls). The suggestion is to translate these into C++ making use of OpenCV calls., so they could be included into OpenCV as IMHO a valuable addition.
^^ broken wiki link
Sorry, link: https://en.wikipedia.org/wiki/Foregro...
yea, that's what your algorithm does not. ;(
Ok, let's keep things constructive. Though as a researcher I have limited time / resources, I am interested to continue to work with the Open Source community (this would like with image tracking project: https://github.com/folterj/BioImageOp...). This is my first contribution on this forum, so just keen to see if there is interest.
so, if you want to add something: