AFTEREFFECTS

News, Information & Workflows from Users & the Adobe Ae Team

bicubic sampling for improved scaling and other transformations in After Effects

In our list of what’s new and changed in After Effects CC (12.0), we mentioned the addition of bicubic sampling for layer transformations, including scaling. Here’s more detail about this new feature.

Beginning in After Effects CC (12.0), you can choose between bicubic and bilinear sampling for selected layers, which determines how pixels are sampled for transformations such as scaling.

This per-layer setting is in the Layer > Quality menu, and it is only relevant for layers with quality set to Best.

The default keyboard shortcuts for setting the sampling method for selected layers are Alt+B (Windows) and Option+B (Mac OS) for Best/Bilinear and Alt+Shift+B (Windows) and Option+Shift+B (Mac OS) for Best/Bicubic.

You can also switch to Draft, Best/Bilinear, and Best/Bicubic for selected layers by clicking the Quality switch in the Switches column of the Timeline panel. When the layer is set to Best quality, the icon shown in the Quality column indicates whether the resampling method is set to bilinear or bicubic.

The bicubic sampling in After Effects should perform better than the related option in Photoshop; the After Effects algorithm preserves over-range and under-range values more consistently and works better (with fewer quantization errors) at extreme scales.

Note that textures in the ray-traced 3D renderer do not use the new bicubic sampling; they always use bilinear sampling. Transformations within effects also still use bilinear sampling, unless the effect specifically implements another method (as with a dedicated scaling plug-in effect or distortion effect).

Bicubic sampling is somewhat more processor-intensive than bilinear sampling, and bicubic sampling is not the highest-quality choice in all cases, so don’t think that you should set it for every layer. It’s rather easy to see artifacts with bicubic sampling in some circumstances, such as ringing and overshoots at a hard transition from one color to another. Bicubic sampling tends to be the best option in cases where transitions from one color to another are more gradual, as is the case with nearly all real-world photographic images, but not necessarily for sharp-edged graphics. Bicubic sampling helps more for scaling up than it does for scaling down.

Here are some examples that show the difference, including one photorealistic example for which bicubic sampling is better and a couple of graphics examples with abrupt color transitions for which bilinear sampling is better:



bicubic sampling, scaled to 800%; better detail and fewer artifacts in pearls and skin



bilinear sampling, scaled to 800%; lost details and worse artifacts in pearls and skin



bicubic sampling of grid layer, camera viewing it edge-on from very close; overshoot artifacts at intersections



bilinear sampling of grid layer, camera viewing it edge-on from very close; with less severe artifacts at intersections



bicubic sampling of grid layer scaled to 1600%; with more severe ringing and overshoot artifacts



bilinear sampling of grid layer scaled to 1600%; with less severe artifacts

TAGGED AS:

Sorry, the comment form is closed at this time.

Join The Discussion