Sunset Lake Software - Comments for "Perspective in GPUImageTransformFilter"

All I do is apply a 4x4

All I do is apply a 4x4 transformation matrix to an image. You're going to need to do the same calculations that these perspective functions apply to create their end matrix. Once you have that matrix, you just feed it into the transformation filter.

Hi Brad, firstly well done on

Hi Brad, firstly well done on GPUImage. Just came across it and looks like it will be really helpful for an app I'm working on.

Building on the issues of the guy above, I'm trying to understand how GPUImage handles the camera position and vanishing point setting in OpenGL. I need to set these very precisely, and I see openGL has methods like gluLookAt but not sure how or if I can use these together with GPUImage, and if not then how does GPUImage handle these issues.

Thanks, Michael.

The math doesn't translate

The math doesn't translate directly across from a CATransform3D used for a Core Animation layer to an OpenGL ES quad, so you have to use slightly different matrix elements for perspective using GPUImage. Look at the 3-D transform example in the FilterShowcase, which applies perspective to a rotating image.

In that case, I needed to set m33 to 0.4 as well as m34 in order to achieve the kind of perspective you expect when doing this in Core Animation.