What About 2-D Grid Corrections?

< Previous | Contents | Manuals Home | Boris FX | Next >

‌What About 2-D Grid Corrections?

When you are correcting distortion on a shot, you may see an asymmetry, where there is more distortion on one side of the shot, and less in another. If you correct one side, the other goes out of whack.

Much of the cause of asymmetry is de-centering—you are correcting the distortion using the wrong optic center, which results in an apparent asymmetry. If you use a grid-type correction, you will likely fix the images, but not the imaging geometry, and the entire match-move will come out wrong. When you fix the centering, the distortion will go away properly without the need for an asymmetric grid-based correction—and the match-move will come out right in the end. SynthEyes fixes de- centering by padding the image, as described in the section on that topic.

You might think or hear about grid-based distortion correction, to rubber-sheet morph the different parts of the image individually into their correct places. This seems a simple approach to the problem, and it is, sort of! But typical industry practices with multiresolution grids are insufficient to do accurate lens correction. They have omissions in the interior, and typically have less and less information as you go to the boundary of the image, exactly when you need MORE information, not less. And many industry grid images don't extend to the edge of the image. As a result it's common for grid images to "burn in" systematic errors in the grid itself.

A model-based approach is preferable, ie one using centering and higher-order terms to better accommodate more complex lenses, such as anamorphic.

That said, Shot/Rectify Lens Grid can be used to process good lens grid images and generate STMaps that “correct’ lens distortion. See the Camera Calibration manual.

< Previous | Contents | Manuals Home | Boris FX | Next >