Resample
Utility node used internally within GafferImage, but not intended to be used directly by end users.
user
Container for user-defined plugs. Nodes should never make their own plugs here, so users are free to do as they wish.
out
The output image generated by this node.
enabled
Turns the node on and off.
in
The input image
matrix
The transform to be applied to the input image. This must contain only translation and scaling.
filter
The filter used to perform the resampling. The name of any OIIO filter may be specified. The default automatically picks an appropriate high-quality filter based on whether or not the image is being enlarged or reduced.
filterScale
A multiplier for the scale of the filter used. Scaling up gives a softer result, scaling down gives a sharper result ( likely to alias or even create black patches where no pixels can be found ). Less than 1 is not recommended unless you have a special technical reason.
boundingMode
The method used when a filter references pixels outside the input data window.
expandDataWindow
Expands the data window by the filter radius, to include the external pixels affected by the filter.
debug
Enables debug output. The HorizontalPass setting outputs an intermediate image filtered just in the horizontal direction - this is an internal optimisation used when filtering with a separable filter. The SinglePass setting forces all filtering to be done in a single pass (as if the filter was non-separable) and can be used for validating the results of the the two-pass (default) approach.
filterDeep
When on, deep images are resized accurately using the same filter as flat images. When off, deep images are resized using the Nearest filter.
Filters with negative lobes ( such as Lanczos3 which is the Default for downscaling ) cannot be represented at all depths with perfect accuracy, because deep alpha must be between 0 and 1, and must be non-decreasing. In extreme cases, involving bright segments with very low alpha, it may be preferable to choose a softer filter without negative lobes ( like Blackman-Harris ).
Caution
When deep images are resized using a filter, many additional deep samples are created per pixel. These slow down subsequent image processing and can create prohibitively large files. These overheads should be mitigated by using a DeepToFlat or DeepHoldout node soon after the Resize.