Last Revision Date: 5/26/2014
This article describes the rectification process of the Ladybug3 camera system.
Rectification is the process of taking a distorted image from one of the 6 Ladybug3 sensors and generating an image from that sensor that fits a pin-hole camera model. This image will have the barrel distortion of the lens removed.
During this process, we wish to ensure that all pixels within the distorted image also appear in the rectified image. This is desirable to minimize the loss of information in the rectification process. Unfortunately, the raw image will not appear rectangular in the rectified image space, and consequently not all pixels in the rectified image will have meaningful image data.
The pixel-by-pixel relationship between the distorted and rectified images can be obtained by the following Ladybug API function calls defined in the ladybuggeom.h file:
You can see an example of a rectified image here: Ladybug3 rectification sample
The lenses used in the Ladybug3 are very high distortion, making them difficult to calibrate precisely. The regions of highest distortion are the corners of the image. In these regions it is common to find the rectification accuracy as bad as approximately 2 pixels. However, in the rest of the rectified image the rectification accuracy is usually within 0.5 pixels. Errors in the corners of rectified images do not impact the panoramic image stitching, as corner pixels are given a low weighting during the stitching process.