Bounding box
Format versus bounding box (bbox), and how nodes grow or clip it.
An image has two rectangles:
- the format: the frame size (for example 1920×1080), what the Viewer frames and what a Write outputs;
- the bounding box (bbox): the area that actually holds pixels. It can be smaller than the format (a small element) or reach outside it (a blur or a move that spills over the edge).
Only the bbox is stored and processed, so a small element in a 4K frame costs little. Coordinates are in pixels, origin bottom left, y up.
How nodes change it
| Node | bbox |
|---|---|
| Read | an EXR's data window; the display window is the format |
| Merge | set bbox to: union (default), intersection, A or B |
| Crop | becomes the crop box; reformat also makes the box the new format |
| Blur | grows by the blur reach; with crop to format on (default) a box that was inside the format stays inside it |
| Reformat | carried through; with preserve bounding box off it is clipped to the new format |
| Transform | moves with the image |
| BlackOutside | outside the bbox is black instead of repeating the edge pixels |
Outside the bbox an image either repeats its edge pixels or is black. Nodes that do not handle the bbox themselves (the colour nodes and Write, for example) see the full format. Merge, Shuffle, Copy, Sharpen, Switch and the other filter and transform nodes work on the bbox directly.