Subject: | HSV : Negative values of the hue |
I have explored HSV values with GD::Simple.
The final version of my exploration script is attached
to this bug report. It checks 5832 RGB triples, executes a
round trip RGB -> HSV -> RGB and compares the initial and final
RGB triples. Of course, there will be rounding errors, so the
final RGB triple will be a little different from the initial.
So I compare the Manhattan distance between the two triples and
compare it with a fuzz value.
The result is that 340 triples out of 5832 give a distance more than 3.
In some cases, the distance is 7. In addition, all these 340 triples
involve a negative value of the hue, although according to the POD,
it should be in the 0..255 interval.
By patching GD::Simple with the attached patch, no triple gives a
result distance more than 3. And all HSV values, including the hue,
are in the 0..255 interval.
In addition, the patch fixes a typo (a copy-paste error) in the POD.
Subject: | patch |
Message body not shown because it is not plain text.
Subject: | rgb-hsv-rgb.pl |
#!/usr/bin/perl
# -*- encoding: utf-8; indent-tabs-mode: nil -*-
use v5.10;
use strict;
use warnings;
use Carp;
use GD::Simple;
my $fmt = " %3d" x 10;
my $step = 15;
my $fuzz = 3;
my $neg_fix = 0;
for (my $r0 = 0; $r0 <= 255; $r0 += $step) {
for (my $g0 = 0; $g0 <= 255; $g0 += $step) {
for (my $b0 = 0; $b0 <= 255; $b0 += $step) {
my ($h, $s, $v) = GD::Simple->RGBtoHSV($r0, $g0, $b0);
my ($r1, $g1, $b1) = GD::Simple->HSVtoRGB($h, $s, $v);
my $delta = abs($r1 - $r0) + abs($g1 - $g0) + abs($b1 - $b0);
if ($delta > $fuzz) {
say sprintf $fmt, $h, $s, $v, $r0, $g0, $b0, $r1, $g1, $b1, $delta;
}
}
}
}