Distortion modeling


If distortion is not specified in blender and I specify the resolution to be 1920x1080 blender renders a scene using a simpele rectilinear pinhole projection and i get a nice rectangular image of 1920x1080. If I add in the compositor a movie distortion node I can render a scene with distortion. However in the case of barrel distortion the 1920x1080 output will have a lot of “black” area because of missing undistorted input image data.

It would be really nice if the output image would be really 1920x1080 and not have those “black” areas. Modelling real camera’s would become easier because real camera’s also don’t have that “black” areas isn’t it?

Has there already been done some work to implement such a feature? Can you provide me withsome pointers to people or url’s that can help in the implemention of such a feature?



You can render overscan – larger than 1920 1080 then there are pixels to fill in the black
when you distort in the compositor. The only problem is the compositor doesn’t have
a concept of the render size, it will work to the size of the first image you add to the node
tree so you would have to force 1920 1080 by loading a background of that size.