Value counts of the picture bands of several rasters (the tiles of one
picture), made to hold up on real imagery. Null when there is no value.
Two things in satellite products spoil a plain rasterHistogram
(its bins spread evenly from the smallest value to the largest):
A fill value that is not tagged as no data (-9999, -32768, the lowest
float, 0 around a scene) takes the corners of the scene. It is often
more than the few percent a stretch clips, so it becomes black (or
white) and the scene itself lands at the other end, all white (or all
black). The smallest or largest value of a band is taken as such a fill
value, and left out, when it lies further from the rest of the values
than the rest spreads (between its 1st and 99th percentile).
A few far outlying values (hot pixels, a fill value on overview edges
blended with the scene) squeeze the values into a handful of bins. The
bins are then taken again over the 1-99 % range widened by its own width
on each side, with the values outside counted in the end bins.
Imagery without either gets the same histogram as rasterHistogram
over the rasters' range.
Value counts of the picture bands of several rasters (the tiles of one picture), made to hold up on real imagery. Null when there is no value.
Two things in satellite products spoil a plain rasterHistogram (its bins spread evenly from the smallest value to the largest):
Imagery without either gets the same histogram as rasterHistogram over the rasters' range.