Barrel Distortion Correction Image Processing Algorithm for Fish-Eye Lenses
- Created: Wednesday, 30 April 2014
In addition to correcting severely distorted fish-eye images, this algorithm finds the optical center of the image, which may be useful for many image processing applications.
John F. Kennedy Space Center, Florida
Immediately following firing of the solid rocket booster (SRB) separation motors during STS-134 ascent, an unknown object resembling a headless bolt was captured by one of the SRB onboard cameras. This object had a length-to-width ratio of approximately 4:1 and appeared to be cylindrical. The end widths appeared slightly wider, giving it the appearance of a bolt-like object. The goal of this investigation was to determine the trajectory and origin of the bolt relative to the ET (external tank) and SRB.
One of the greatest challenges of this investigation was correcting the fish-eye images acquired from the wide-angle lens used with this camera. The first step in photogrammetry-based trajectory analysis is to transform any geometric distortion to an equivalent ideal undistorted image. Standard distortion algorithms were not able to produce images suitable for refined trajectory analysis sufficient to extrapolate the origin of the debris in a 3D CAD (computer-aided design) model to locate the origin.
Debris photogrammetry was performed on the SRB bolt images. After correcting the SRB camera lens distortion with a custom algorithm implemented for this specific set of images, a family of trajectories was computed based on a family of assumed object sizes. Since previous image correction attempts fell short of what was needed to accomplish the photogrammetry trajectory analysis, a software algorithm was customized for this particular problem.
Software based on a previous project was utilized. The code was written in FORTRAN and was modified to include additional correction parameters for this specific problem. The original algorithm used a single parameter to correct quadratic distortion, from an image where the optical axis of the lens was coincident with the center of the image. In this new implementation, two additional parameters were added. These parameters made the quadratic strength of the correction a function of radial distance. Another challenge in the image correction, known from the start, was that the optical axis of the lens was very much off-center from the image center. The information on orientation was lost when the camera was removed from the retrieved SRB. To solve the problem, a search algorithm was developed and coded to find the best-fit optical center of the image, as well as the best-fit correction terms.
As compared to previous correction methods, this algorithm seemed to outperform standard methods by taking a more aggressive approach and forcing the output to produce truly straight lines where straight lines were known to exist. In addition to correcting severely distorted fish-eye images, this algorithm finds the optical center of the image, which may be useful for many image processing applications.
This work was done by John Lane of EASI for Kennedy Space Center. For more information, contact the Kennedy Space Center Technology Transfer Office at 321-867-5033. KSC-13751.