Hello,
I'm working with the Axis M3007-PV network camera. It has a horizontal angle of view of 187°, vertical of 168° and produces fisheye-like images. I'm trying to undistort/dewarp these images so that straight lines look straight again; I'm not trying to unwrap the image into a panoramic a la Golan's Bloggie unwrap.
So far I've tried two techniques: chessboard calibration/undistort using ofxCv, and the less flexible approach of producing a spherical (partial) map.
With chessboard/undistort, the results vary wildly but I have had some success. The main issues occur around the edge of the image. This is based on adding around 10 calibration images with the chessboard at various positions:
With spherical mapping, the centre looks ok but the edges are heavily distorted:
And so to the question: does anyone have any experience with either/both approaches? Is there a way to improve either or is there an alternative approach that I may have missed?
It's for an arts/theatre project BTW, and all code will be open-sourced once it's complete!
Cheers