Usage guide for Compositing with Carbon Basic Deformation Filter
The Deformation filter takes a key created by key processing and produces a mesh
to be used by the blend function during the resizing.
The Deformation filter has three types of channels, Object, Subject, and Alpha.
The object channel data is a "bump", or density, map. This channel gets its name from the part of the transform it
acts on, it controls what goes into the object of the transform. The higher the value for any given pixel in the
key, the greater the area of the image used to compose the resultant pixel. Key value of zero results in a
transparent pixel in the result. For demonstration purposes patterns were used in the following images. However
images and generated shapes may be used as well. Use low values, scaled if needed, to avoid extreme deformation.
The subject channel affects the opposite end of the transform. The higher the value of a key pixel, the greater
the subject pixel is spread out over the resultant image. A key value of zero excludes its corresponding image
pixel from the result.
The alpha channel is used to refine the edge of a deformation for anti-aliasing. It acts the same as the alpha
filter itself, but is included with the deformation filter to eliminate the need for two filters for most uses.
The deformation filter has two methods for handling the edges when transparent pixels are created. "Alternate Edge"
toggles between these methods.
"Radial Smoothing" can improve most deformations, and is usually engaged. However, you may wish to disable it
for some deformations. It is a single pass algorithm designed to reduce variations in density between adjacent
pixels along radial lines from the lowest density area in all directions. The algorithm is not applied to meshes
that have no detectable low density area.