Heal transparent

Replaces transparent pixels with good candidate pixels found in the neighborhood of the missing pixel.
seek radius
Maximum distance in neighborhood we look for better candidates per improvement.
name: seek-distance type: int default: 10 minimum: 1 maximum: 512 ui-minimum: 1 ui-maximum: 512 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 100
seek reduction
factor seek distance is shortened, until we're about 2px short per iteration, 1.0 means no reduction
name: seek-reduction type: double default: 0.60 minimum: 0.00 maximum: 1.00 ui-minimum: 0.00 ui-maximum: 1.00 ui-gamma: 1.00 ui-step-small: 0.00 ui-step-big: 0.10 ui-digits: 3
min rounds
Ensuring that we get results even with low retry chance
name: min-iter type: int default: 20 minimum: 1 maximum: 512 ui-minimum: 1 ui-maximum: 512 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 100
min neighbors
minimum neighbors that must be set before we consider setting
name: min-neighbors type: int default: 1 minimum: 0 maximum: 4 ui-minimum: 0 ui-maximum: 4 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 2
max rounds
Mostly a saftey valve, so that we terminate
name: max-iter type: int default: 800 minimum: 1 maximum: 40000 ui-minimum: 1 ui-maximum: 40000 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 5
iterations per round per probe
number of improvement iterations, after initial search - that each probe gets.
name: iterations type: int default: 23 minimum: 1 maximum: 1000 ui-minimum: 1 ui-maximum: 1000 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 100
histogram
only consider pixels whose neighborhood color distribution is smaller than specified, set to 1.0 for much faster processing, but needing lower seek distance to yield ok results.
name: histogram type: double default: 0.50 minimum: 0.00 maximum: 1.00 ui-minimum: 0.00 ui-maximum: 1.00 ui-gamma: 1.00 ui-step-small: 0.00 ui-step-big: 0.10 ui-digits: 3
ring gap1
radius, in pixels of nearest to pixel circle of neighborhood metric
name: ring-gap1 type: double default: 1.10 minimum: 0.00 maximum: 16.00 ui-minimum: 0.00 ui-maximum: 16.00 ui-gamma: 1.00 ui-step-small: 0.25 ui-step-big: 0.25 ui-digits: 3
ring gap2
radius, in pixels of second nearest to pixel circle
name: ring-gap2 type: double default: 1.90 minimum: 0.00 maximum: 16.00 ui-minimum: 0.00 ui-maximum: 16.00 ui-gamma: 1.00 ui-step-small: 0.25 ui-step-big: 0.25 ui-digits: 3
ring gap3
radius, in pixels of third pixel circle
name: ring-gap3 type: double default: 3.20 minimum: 0.00 maximum: 16.00 ui-minimum: 0.00 ui-maximum: 16.00 ui-gamma: 1.00 ui-step-small: 0.25 ui-step-big: 0.25 ui-digits: 3
ring gap4
radius, in pixels of fourth pixel circle (not always in use)
name: ring-gap4 type: double default: 4.10 minimum: 0.00 maximum: 16.00 ui-minimum: 0.00 ui-maximum: 16.00 ui-gamma: 1.00 ui-step-small: 0.25 ui-step-big: 0.25 ui-digits: 3
scale needles
name: scale-needles type: int default: 5 minimum: 1 maximum: 7 ui-minimum: 1 ui-maximum: 7 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 5
rounds
number of improvement iterations, after initial search - that each probe gets.
name: rounds type: int default: 16 minimum: 0 maximum: 1000 ui-minimum: 0 ui-maximum: 1000 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 100
try probability
The chance that a candidate pixel probe will start being filled in
name: chance-try type: double default: 0.01 minimum: 0.00 maximum: 1.00 ui-minimum: 0.00 ui-maximum: 1.00 ui-gamma: 1.00 ui-step-small: 0.01 ui-step-big: 0.10 ui-digits: 3
retry probability
The chance that a pixel probe gets an improvement in an iteration
name: chance-retry type: double default: 0.90 minimum: 0.00 maximum: 1.00 ui-minimum: 0.00 ui-maximum: 1.00 ui-gamma: 1.00 ui-step-small: 0.01 ui-step-big: 0.10 ui-digits: 3
chance neighbor
The chance of trying to find neighbor improvements
name: chance-neighbor type: double default: 1.00 minimum: 0.00 maximum: 1.00 ui-minimum: 0.00 ui-maximum: 1.00 ui-gamma: 1.00 ui-step-small: 0.01 ui-step-big: 0.10 ui-digits: 3
dist powk
influences the (lack of) importance of further away pixels
name: metric-dist-powk type: double default: 1.80 minimum: 0.00 maximum: 10.00 ui-minimum: 0.00 ui-maximum: 10.00 ui-gamma: 1.00 ui-step-small: 0.10 ui-step-big: 1.00 ui-digits: 3
empty hay score
score given to pixels that are empty, in the search neighborhood of pixel, this being at default or higher value sometimes discourages some of the good very nearby matches
name: metric-empty-hay-score type: double default: 0.28 minimum: 0.00 maximum: 100.00 ui-minimum: 0.00 ui-maximum: 100.00 ui-gamma: 1.00 ui-step-small: 0.05 ui-step-big: 0.10 ui-digits: 2
empty needle score
the score given in the metric to an empty spot
name: metric-empty-needle-score type: double default: 0.18 minimum: 0.00 maximum: 100.00 ui-minimum: 0.00 ui-maximum: 100.00 ui-gamma: 1.00 ui-step-small: 0.05 ui-step-big: 0.10 ui-digits: 2
metric cohesion
influences the importance of probe spatial proximity
name: metric-cohesion type: double default: 0.13 minimum: 0.00 maximum: 100.00 ui-minimum: 0.00 ui-maximum: 100.00 ui-gamma: 1.00 ui-step-small: 0.20 ui-step-big: 0.20 ui-digits: 2
ring twist
incremental twist per circle, in radians
name: ring-twist type: double default: 0.00 minimum: 0.00 maximum: 1.00 ui-minimum: 0.00 ui-maximum: 1.00 ui-gamma: 1.00 ui-step-small: 0.01 ui-step-big: 0.20 ui-digits: 3
direction invariant
wheter we normalize feature vector to start with highest energy ray
name: direction-invariant type: boolean default: True
source neighbors
pick neighbor of neighbors as starting point if good, 4connected 8conntected or 12\/16 with longer teleport
name: source-neighbors type: int default: 8 minimum: 0 maximum: 16 ui-minimum: 0 ui-maximum: 16 ui-gamma: 1.00 ui-step-small: 1 ui-step-big: 5
name: gegl:alpha-inpaint
pads: input output
parent-class: GeglOperationFilter
categories: heal
operation-class:  ;)
GEGL This page is part of the online GEGL Documentation, GEGL is a data flow based image processing library/framework, made to fuel GIMPs high-bit depth non-destructive editing future.