Comparing computer vision libraries in python [closed]

Question

I want to decide about a Python computer vision library. I had used OpenCV in C++, and like it very much. However this time I need to develop my algorithm in Python. My short list has three libraries: 1- OpenCV (Python wrapper) 2- PIL (Python Image Processing Library) 3- scikit-image

Would you please help me to compare these libraries?

I use numpy, scipy, scikit-learn in the rest of my code. The performance and ease of use is an important factor, also, portability is an important factor for me.

Thanks for your help

Answer 1

I have worked mainly with OpenCV and also with scikit-image. I would say that while OpenCV is more focus on computer vision (classification, feature detection and extraction,...). However lately scikit-image is improving rapidly. I faced that some algorithms perform faster under OpenCV, however in most cases I find much more easier working with scikit-image, OpenCV documentations is quite cryptic. As long as OpenCV 2.x bindings works with numpy as well as scikit-image I would take into account using both libraries, trying to take the better of each of them. At least is what I have done in my last project.