smoothing op• Data kinds: image → image
• Call: fullseye.apply(img, "xkor_unsharp", a=0.5, b=0.5) (the 2-D model is one image plus two scalar knobs a,b∈[0,1])
• GPU: none (CPU reference implementation only)
*The figure is the actual output on a synthetic 128×128 input. Left: input, right: output. Point clouds are drawn as a top-down scatter (brightness = z), 1-D series as a line plot, volumes as the maximum-intensity projection along z, videos as the middle frame, complex images as magnitude; return values that are not pictures are shown as the values themselves.*
*Knob a does not change the output (measured: identical at 0.1 / 0.5 / 0.9).*
Sweeping knob b (0.1 / 0.5 / 0.9, the other knob at its default):
▸ xkor_unsharp: knob b sweep (docs site)
On other images (synthetic scene / photo / coins. Top row: inputs, bottom row: their outputs. Knobs at default):
▸ xkor_unsharp: other inputs (docs site)
*The 4th column is a colour (H,W,3) input. This op handles colour without crossing channels (it does not convolve the colour channel as a third spatial axis).*
Unsharp mask (sharpening). Calls kornia.filters.unsharp_mask with a window sized to sigma (v + (v - G_σ * v), strength fixed at 1).
a is unused; b sweeps the blur sigma (0.5 + 2.0 * b, range 0.5–2.5). The classic sharpening that subtracts the low-frequency (blurred) component and adds the emphasised difference back. The window radius is int(4σ + 0.5) and the border mirrors including the edge pixel (as in xkor_gaussian). The knobs mean something different from the core unsharp: there a is the strength (1.5a) and b is σ = 0.5 + 1.5b. The core unsharp(v, 2/3, (s - 0.5)/1.5) and this op with b = (s - 0.5)/2 are the same operation.
★Fix of 2026-11-03 (values change): until then the window was a fixed 5x5, so even σ = 1.5 (b = 0.5) was cut at radius 2 (the same defect as xkor_gaussian).
Other implementations of the same operation (measured; docs/IMPL_DIFFERENTIAL.md): on the same inputs and the same a, b, the interior difference (relative to the value range; below 5e-3 counts as a match) is unsharp 0.19. Why the mismatching ones differ (knob mapping, window, border, quantisation, a different definition) is in the table.
• gallery2d_smoothing_rank family guide
• 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.
The program below has been verified to run (same input as the figure). In Studio's help this block becomes buttons that load and run it on the spot.
xkor_unsharp 0.35 0.50
▸ Load this pipeline · Load & run
• gallery2d_smoothing_rank — py -3.11 examples/gallery2d_smoothing_rank.py
image as input)identity · gaussian · mean_box · mean_box_wh · bilateral · kuwahara · unsharp · median
smoothing)gaussian · mean_box · mean_box_wh · bilateral · kuwahara · unsharp · sk_tv · sk_wavelet
*Provenance: ops.py — 2D 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.