VSE: Speedup Subsampled 3x3 image filter #117125

Aras Pranckevicius · 2024-01-15T10:50:09+01:00

2024-01-15 10:50:09 +01:00

Make Subsampling 3x3 filter twice faster (on 4K UHD resolution, Windows/VS2022/Ryzen5950X: 52.7ms -> 28.3ms), by reformulating how it works:

Conceptually Subsampling filter is a box filter: it sums up N source image pixels, computes their average and outputs the result. Critical
thing is, that should be done in premultiplied space so that colors from fully or mostly transparent regions do not "override" opaque colors.

Previously, when operating on byte images, the code achieved this by always working on byte values, doing "progressively smaller"
lerp into byte color result, taking care of premultiplication and again storing the "straight" alpha for each sample being processed.
This meant that for each sample, there are 3 divisions involved! This also led to some precision loss, since for all 9 samples all the intermediate results would only be stored at byte precision.

Reformulate that by simply accumulating the premultiplied color as a float. This gets rid of all divisions, except the last step when said float needs to be written back into a byte color.

The unit test results have a tiny difference, since now it is arguably better (as per above, previously it was having some precision loss).

Make Subsampling 3x3 filter twice faster (on 4K UHD resolution, Windows/VS2022/Ryzen5950X: 52.7ms -> 28.3ms), by reformulating how it works: Conceptually Subsampling filter is a box filter: it sums up N source image pixels, computes their average and outputs the result. Critical thing is, that should be done in premultiplied space so that colors from fully or mostly transparent regions do not "override" opaque colors. Previously, when operating on byte images, the code achieved this by always working on byte values, doing "progressively smaller" lerp into byte color result, taking care of premultiplication and again storing the "straight" alpha for each sample being processed. This meant that for each sample, there are 3 divisions involved! This also led to some precision loss, since for all 9 samples all the intermediate results would only be stored at byte precision. Reformulate that by simply accumulating the premultiplied color as a float. This gets rid of all divisions, except the last step when said float needs to be written back into a byte color. The unit test results have a tiny difference, since now it is arguably better (as per above, previously it was having some precision loss).

🚀 3

Aras Pranckevicius changed title from ~~WIP: VSE: Optimize Subsampling 3x3 image filter~~ to WIP: VSE: Optimize Subsampling 3x3 filter and simplify code

2024-01-15 10:50:48 +01:00

Aras Pranckevicius force-pushed vse-subsampling from 29889a28df to 2190b707a9

2024-01-16 21:03:23 +01:00

Compare

Aras Pranckevicius changed title from ~~WIP: VSE: Optimize Subsampling 3x3 filter and simplify code~~ to WIP: VSE: Speedup Subsampled 3x3 filter

2024-01-16 21:03:47 +01:00

Aras Pranckevicius force-pushed vse-subsampling from 2190b707a9 to cf619dd743

2024-01-16 21:09:57 +01:00

Compare

Aras Pranckevicius changed title from ~~WIP: VSE: Speedup Subsampled 3x3 filter~~ to VSE: Speedup Subsampled 3x3 image filter

2024-01-16 21:11:10 +01:00

Aras Pranckevicius force-pushed vse-subsampling from cf619dd743 to 724f0c2c9e

2024-01-16 21:18:50 +01:00

Compare

Aras Pranckevicius commented

2024-01-16 21:23:52 +01:00

@blender-bot build