OpenCV 2.0 and Beyond
With our recent post highlighting the work of OpenCV's Victor Ehruhimov on the Texas Robot, it's worth bringing the community up-to-date on the the work that is going into OpenCV.
OpenCV is a computer vision library started as an open source project at Intel. In the last two years, Willow Garage has become the primary backer of this library, which is released under the BSD license, which makes it free for any use, commercial, research, or otherwise. Currently, OpenCV has an active user group of over 40,000 members and has been downloaded well over two million times.
Willow Garage is focused on expanding the capabilities of OpenCV for robot perception. We want OpenCV to be an important library for helping robots see. Before we could do this, we recognized that it was necessary to modernize the decade-old library. OpenCV has undergone a Renaissance, adding full new C++ (by Vadim Pisarevsky) and Python (by James Bowman) interfaces. It has also added new texture recognition features and Marius Muja and Professor David Lowe's Fast Approximate Nearest Neighbor machine learning library (FLANN). There is even a full calibration→rectification→stereo correspondence→depth map pipeline. All of this work culminated with the release of OpenCV 2.0 in September 2009.
With this new, modern library, the OpenCV team is now moving forward on implementing new features as well as making the library more stable and tested than ever before. You can follow the increasing test code coverage, or, if you prefer features, you can follow the OpenCV developer Twitter account.
-- Gary Bradski