version 1, including all changes.
.
Rev |
Author |
# |
Line |
1 |
perry |
1 |
pnmgamma |
|
|
2 |
!!!pnmgamma |
|
|
3 |
NAME |
|
|
4 |
SYNOPSIS |
|
|
5 |
DESCRIPTION |
|
|
6 |
WHAT IS GAMMA? |
|
|
7 |
SEE ALSO |
|
|
8 |
AUTHOR |
|
|
9 |
---- |
|
|
10 |
!!NAME |
|
|
11 |
|
|
|
12 |
|
|
|
13 |
pnmgamma - perform gamma correction on a PNM image |
|
|
14 |
!!SYNOPSIS |
|
|
15 |
|
|
|
16 |
|
|
|
17 |
__pnmgamma__ [[__-ungamma__] [[__-cieramp__] |
|
|
18 |
[[''value'' [[''pnmfile'']]__ |
|
|
19 |
pnmgamma__ [[__-ungamma__] [[__-cieramp__] ''redgamma |
|
|
20 |
greengamma bluegamma'' [[''pnmfile''] |
|
|
21 |
!!DESCRIPTION |
|
|
22 |
|
|
|
23 |
|
|
|
24 |
Reads a psuedo-PNM image as input. Performs gamma |
|
|
25 |
correction, and produces a PNM image as output. |
|
|
26 |
|
|
|
27 |
|
|
|
28 |
Alternatively, this program can undo gamma |
|
|
29 |
correction. |
|
|
30 |
|
|
|
31 |
|
|
|
32 |
The PPM specification states that the sample values in the |
|
|
33 |
image are nonlinear, i.e. not directly proportional to light |
|
|
34 |
intensity, i.e. gamma corrected. But there exist images that |
|
|
35 |
are PPM in every respect except that their sample values are |
|
|
36 |
directly proportional to light intensity. People may loosely |
|
|
37 |
refer to these as PPM, but they are not. __pnmgamma__ |
|
|
38 |
converts these pseudo-PPM images to true PPM by performing |
|
|
39 |
gamma correction. To get true PPM, you must specify the |
|
|
40 |
__-cieramp__ option and no gamma values. That causes |
|
|
41 |
__pnmgamma__ to apply the CIE Rec. 709 gamma transfer |
|
|
42 |
function, as specified by the PPM format |
|
|
43 |
specification. |
|
|
44 |
|
|
|
45 |
|
|
|
46 |
On the other hand, you can use the __-ungamma__ option to |
|
|
47 |
convert from true PPM to linear RGB pseudo-PPM. (Again, if |
|
|
48 |
the input is true PPM, specify the __-cieramp__ option |
|
|
49 |
and no gamma values). |
|
|
50 |
|
|
|
51 |
|
|
|
52 |
The situation for PGM images is analogous. And |
|
|
53 |
__pnmgamma__ treats PBM images as PGM |
|
|
54 |
images. |
|
|
55 |
|
|
|
56 |
|
|
|
57 |
You can also apply a different transfer function (which |
|
|
58 |
means you don't end up with a true PPM image) by selecting |
|
|
59 |
the gamma values as arguments or omitting the |
|
|
60 |
__-cieramp__ option. The gamma value is the power to |
|
|
61 |
which the input value is raised in the transfer function. A |
|
|
62 |
value of 1 means the output is the same as the input. A |
|
|
63 |
value less than one makes the output samples numerically |
|
|
64 |
less than the input samples; A value greater than one makes |
|
|
65 |
the samples numerically greater. |
|
|
66 |
|
|
|
67 |
|
|
|
68 |
Without the __-cieramp__ option, the transfer function is |
|
|
69 |
a simple power function. With __-cieramp__, it is a power |
|
|
70 |
function modified with a linear ramp near black, as |
|
|
71 |
described in CIE Rec. 709. |
|
|
72 |
|
|
|
73 |
|
|
|
74 |
When you feed a linear PPM image to a display program that |
|
|
75 |
expects a true PPM, the display appears darker than it |
|
|
76 |
should, so __pnmgamma__ has the effect of lightening the |
|
|
77 |
image. When you feed a true PPM to a display program that |
|
|
78 |
expects linear sample values, and therefore does a gamma |
|
|
79 |
correction of its own on them, the display appears lighter |
|
|
80 |
than it should, so __pnmgamma__ with a gamma value less |
|
|
81 |
than one (the multiplicative inverse of whatever gamma value |
|
|
82 |
the display program uses) has the effect of darkening the |
|
|
83 |
image. |
|
|
84 |
!!WHAT IS GAMMA? |
|
|
85 |
|
|
|
86 |
|
|
|
87 |
A good explanation of gamma is in Charles Poynton's GammaFAQ |
|
|
88 |
at |
|
|
89 |
|
|
|
90 |
|
|
|
91 |
In brief: The simplest way to code an image is by using |
|
|
92 |
sample values that are directly proportional to the |
|
|
93 |
intensity of the color components. But that wastes the |
|
|
94 |
sample space because the human eye can't discern differences |
|
|
95 |
between low-intensity colors as well as it can between |
|
|
96 |
high-intensity colors. So instead, we pass the light |
|
|
97 |
intensity values through a transfer function that makes it |
|
|
98 |
so that changing a sample value by 1 causes the same level |
|
|
99 |
of perceived color change anywhere in the sample range. We |
|
|
100 |
store those resulting values in the image file. That |
|
|
101 |
transfer function is called the gamma transfer function and |
|
|
102 |
the transformation is called gamma correcting. |
|
|
103 |
|
|
|
104 |
|
|
|
105 |
Virtually all image formats, either specified or de facto, |
|
|
106 |
use gamma-corrected values for their sample |
|
|
107 |
values. |
|
|
108 |
|
|
|
109 |
|
|
|
110 |
What's really nice about gamma is that by coincidence, the |
|
|
111 |
inverse function that you have to do to convert the |
|
|
112 |
gamma-corrected values back to real light intensities is |
|
|
113 |
done automatically by CRTs. You just apply a voltage to the |
|
|
114 |
CRT's electron gun that is proportional to the |
|
|
115 |
gamma-corrected sample value, and the intensity of light |
|
|
116 |
that comes of the screen is close to the intensity value you |
|
|
117 |
had before you applied the gamma transfer |
|
|
118 |
function! |
|
|
119 |
|
|
|
120 |
|
|
|
121 |
And when you consider that computer video devices usually |
|
|
122 |
want you to store in video memory a value proportional to |
|
|
123 |
the signal voltage you want to go to the monitor, which the |
|
|
124 |
monitor turns into a proportional drive voltage on the |
|
|
125 |
electron gun, it is really convenient to work with |
|
|
126 |
gamma-corrected sample values. |
|
|
127 |
!!SEE ALSO |
|
|
128 |
|
|
|
129 |
|
|
|
130 |
pnm(5) |
|
|
131 |
!!AUTHOR |
|
|
132 |
|
|
|
133 |
|
|
|
134 |
Copyright (C) 1991 by Bill Davidson and Jef |
|
|
135 |
Poskanzer. |
|
|
136 |
---- |