post op• Data kinds: rgb → rgb
• Call: import gfx2d; gfx2d.bloom(rgb, threshold=0.8, sigma=4.0, intensity=0.6) (or opsgfx2d.get("bloom"))
Bleed the bright parts of an image into their neighbourhood.
The classical three steps: isolate what is above `threshold`, blur it by a
Gaussian of `sigma, add intensity` times the result back, clip.
A *linear-light* operator: light spreading in a lens adds radiance. Running
it on sRGB-encoded values makes the halo too strong in the mid-tones —
`srgb_to_linear first, linear_to_srgb` after.
Exact identities: `intensity=0 and threshold >= 1` both return the
input bit for bit, and a black image stays black.
The added energy is *not* conserved by the clip at the end. The suite
measures both halves: on a centred bright blob the Gaussian preserves the
bright mass to 5.0e-16 relative, so everything lost after that is the
clip — on a random scene with a saturated square, 2.69 % of the pixels clip
and 1.65 % of the total energy is thrown away, and neither number is
returned to the caller.
• Sample-data catalog (download URLs / licences) — 2-D uses skimage.data (BSD/public domain) plus synthetic images; 3-D lists download URLs for real data sources (Stanford, PDS, …).
• Operator provenance and references — the sources of the research/methods this op family came from.
• The canonical algorithm (author, year) and its uses are named in the family usage guide above.
• gfx2d_scene — py -3.11 examples/gfx2d_scene.py
rgb as input)srgb_to_linear · linear_to_srgb · blend_mode · light_mask · normal_map_decode · vignette · chromatic_aberration · film_grain
post)vignette · chromatic_aberration · film_grain · color_lut · color_grade · dither · palette_quantize
*Provenance: gfx2d.py — GFX2D operator registry. This per-op note is generated by tools/opdocs.py md (do not hand-edit).*
© 2026 Kazufumi Furuse — Fullseye operator documentation. Licensed under Apache-2.0.