1 |
.\" RCSid "$Id: macbethcal.1,v 1.2 2003/12/09 15:59:06 greg Exp $" |
2 |
.TH MACBETHCAL 1 1/16/97 RADIANCE |
3 |
.SH NAME |
4 |
macbethcal - compute color compensation based on measured Macbeth chart |
5 |
.SH SYNOPSIS |
6 |
.B macbethcal |
7 |
[ |
8 |
.B "\-d debug.pic" |
9 |
][ |
10 |
.B "\-p xul yul xur yur xll yll xlr ylr" |
11 |
] |
12 |
.B scannedin.pic |
13 |
[ |
14 |
.B calibout.cal |
15 |
] |
16 |
.br |
17 |
.B macbethcal |
18 |
.B \-c |
19 |
[ |
20 |
.B "\-d debug.pic" |
21 |
] |
22 |
[ |
23 |
.B measured.xyY |
24 |
[ |
25 |
.B calibout.cal |
26 |
] |
27 |
] |
28 |
.SH DESCRIPTION |
29 |
.I Macbethcal |
30 |
takes a scanned image or measurement set |
31 |
of a Macbeth ColorChecker\u\s-3TM\s+3\d color |
32 |
rendition chart and computes a color mapping |
33 |
function suitable as input to |
34 |
.I pcomb(1). |
35 |
.PP |
36 |
In the first form, |
37 |
.I macbethcal |
38 |
takes a scanned image of a Macbeth chart that has been converted |
39 |
into a Radiance picture using a fixed procedure. |
40 |
When used properly as input to |
41 |
.I pcomb, |
42 |
the computed calibration file will adjust the |
43 |
brightness and color of any similarly scanned and converted image |
44 |
so as to best match the original. |
45 |
If the lighting conditions are carefully controlled (as in the case |
46 |
of a flatbed scanner), it is even possible to get reliable |
47 |
reflectance values this way, at least within 10% or so. |
48 |
The input picture is named on the command line. |
49 |
The output calibration file will be written to the standard output |
50 |
if no file name is given on the command line. |
51 |
.PP |
52 |
In the second form, the input is from a file containing measured |
53 |
values for each Macbeth color. |
54 |
This file must contain entries of the form: |
55 |
.sp |
56 |
.nf |
57 |
N x y Y |
58 |
.fi |
59 |
.sp |
60 |
Where |
61 |
.I N |
62 |
is the number of the corresponding Macbeth color. |
63 |
(See back of ColorChecker chart |
64 |
for color names and indexing, but it basically starts from the upper |
65 |
left with 1 and proceeds in English text order to the lower right, |
66 |
which is 24.) |
67 |
The values |
68 |
.I x, |
69 |
.I y |
70 |
and |
71 |
.I Y |
72 |
are the 1931 CIE (x,y) chromaticity coordinates followed by the |
73 |
luminance for that color, which can be in any units. |
74 |
If a white value is known (i.e. maximum output level), then it may |
75 |
be given as entry number 0. |
76 |
The entries may be in any order, and comments may be included |
77 |
delimited by a pound sign ('#') and continuing to the end of line. |
78 |
It is recommended that measurements be done for all 24 colors, |
79 |
but the only required entries are the 6 neutral values on the bottom |
80 |
row of the chart. |
81 |
.PP |
82 |
Computing a mapping from measured colors is usually more convenient |
83 |
when calibrating a particular output device. |
84 |
This is accomplished by printing the picture |
85 |
.I macbeth_spec.pic |
86 |
(which may be found in the standard RADIANCE library directory in |
87 |
the lib subdirectory) and measuring the output with a chroma meter |
88 |
or spectrophotometer. |
89 |
.PP |
90 |
For a scanned image, the locations of the 24 Macbeth patches |
91 |
in the input picture must be known. |
92 |
If the chart borders are not at the edges of the input picture, or |
93 |
the chart has been reversed or rotated or is uncentered or at an |
94 |
oblique angle, then it is necessary to specify the pixel locations |
95 |
of the corners of the chart with the |
96 |
.I \-p |
97 |
option. |
98 |
The corner postions (x,y pixel addresses as given by the |
99 |
.I ximage(1) |
100 |
"p" command) |
101 |
are ordered on the command line: |
102 |
upper-left, upper-right, lower-left, lower-right |
103 |
(i.e. English text ordering). |
104 |
These coordinates should be the outside corner positions of |
105 |
the following patches: |
106 |
.sp |
107 |
.nf |
108 |
upper-left = 1. dark skin |
109 |
upper-right = 6. bluish green |
110 |
lower-left = 19. white |
111 |
lower-right = 24. black |
112 |
.fi |
113 |
.sp |
114 |
If the chart has been flipped or rotated, simply give the pixel |
115 |
positions of the appropriate patch corners, wherever |
116 |
they are in the image. |
117 |
(Note: if the Radiance picture has been flipped or rotated with |
118 |
.I pflip(1) |
119 |
or |
120 |
.I protate(1), |
121 |
.I ximage |
122 |
will report the original pixel positions if the |
123 |
.I \-c |
124 |
option was not used by the reorienting program(s). |
125 |
This will be wrong, so be sure to use the |
126 |
.I \-c |
127 |
option.) |
128 |
.I Macbethcal |
129 |
can handle a chart with any orientation or perspective warping if |
130 |
the corner coordinates are given correctly. |
131 |
The debug picture output is the best way to check for consistency. |
132 |
(See the |
133 |
.I \-d |
134 |
option, below.) |
135 |
.PP |
136 |
The |
137 |
.I \-d |
138 |
option may be used to specify an additional output file, which will |
139 |
be a picture comparing the scanned image processed according to the |
140 |
computed mapping against the standard Macbeth colors. |
141 |
It is a good idea to use the debug option to check that the color |
142 |
patches are being located correctly, and to see how well |
143 |
.I macbethcal |
144 |
does at matching colors. |
145 |
The center of each patch will show the target color; |
146 |
the left side of each patch will show the original |
147 |
color, and the right side will show the corrected value. |
148 |
If the match works well, the debug picture should have a sort of |
149 |
"notch on the left" look in each patch. |
150 |
Macbeth colors that could not be matched because they were out of |
151 |
gamut on this device are indicated with diagonal lines drawn |
152 |
through the associated target colors. |
153 |
.SH METHOD |
154 |
.I Macbethcal |
155 |
computes the color mapping in two stages. |
156 |
The first stage uses the six neutral color patches at the bottom of |
157 |
the Macbeth chart to compute a piecewise linear approximation to the |
158 |
brightness mapping of each RGB primary. |
159 |
The second stage looks at all the colors that are within the |
160 |
device's gamut to compute a |
161 |
least-sqaures fit for a linear color transformation from the |
162 |
measured space into the standard Radiance RGB space (as defined by |
163 |
the three primaries in src/common/color.h). |
164 |
.PP |
165 |
Thanks to the nature of inverse mappings, this method should work |
166 |
either for converting scanned data to match the original, or for |
167 |
preconditioning pictures to be sent to specific output devices. |
168 |
In other words, the same calibration file works either for |
169 |
correcting scanned images OR precorrecting images before printing. |
170 |
.PP |
171 |
A warning is printed if some unsaturated colors are determined to |
172 |
be out of gamut, as this may indicate a poor rendition or improper |
173 |
picture alignment. |
174 |
The debug picture will show which colors were excluded by drawing |
175 |
diagonal lines through their entries. |
176 |
.SH NOTE |
177 |
It is very important that the same settings be applied when |
178 |
scanning or printing other images to be calibrated with the computed file. |
179 |
In particular, all exposure adjustments should be fixed manually, |
180 |
and no tweaking of the settings should be done along the way. |
181 |
The final result will be best if the original scanned image is not |
182 |
too far off from what it should be. |
183 |
In the case of slide and negative scanners, it is best to apply the |
184 |
recommended calibration file for the type of film used, so long as |
185 |
this calibration is fixed and not adjusted on a per-image basis. |
186 |
.SH CHART AVAILABILITY |
187 |
The Macbeth chart is available at most photographic supply stores, |
188 |
or may be ordered directly from Macbeth: |
189 |
.sp |
190 |
.nf |
191 |
Macbeth |
192 |
Munsell Color |
193 |
405 Little Britain Rd. |
194 |
New Windsor, NY 12553-6148 |
195 |
tel. 1-800-622-2384 (USA) |
196 |
fax. 1-914-561-0267 |
197 |
.fi |
198 |
.sp |
199 |
The chart sells for under $50 US at the time of this writing. |
200 |
.SH EXAMPLES |
201 |
To compute a calibration for a FunkyThing scanner and check the |
202 |
results: |
203 |
.IP "" .2i |
204 |
ra_tiff -r mbscan.tif mbscan.pic |
205 |
.br |
206 |
macbethcal -d debug.pic mbscan.pic FunkyThing.cal |
207 |
.br |
208 |
ximage debug.pic |
209 |
.PP |
210 |
To apply this computed calibration to another scanned image: |
211 |
.IP "" .2i |
212 |
ra_tiff -r another.tif | pcomb -f FunkyThing.cal - > another_calib.pic |
213 |
.PP |
214 |
To compute a calibration file for the BigWhiz film recorder, after |
215 |
taking measurements of a slide made from macbeth_spec.pic: |
216 |
.IP "" .2i |
217 |
macbethcal -c macbeth_spec.xyY BigWhiz.cal |
218 |
.PP |
219 |
To prepare a picture prior to output on the same film recorder: |
220 |
.IP "" .2i |
221 |
pcomb -f BigWhiz.cal standard.pic > toprint.pic |
222 |
.PP |
223 |
To use |
224 |
.I pcond(1) |
225 |
to also adjust the image for human response: |
226 |
.IP "" .2i |
227 |
pcond -f BigWhiz.cal -h standard.pic > toprint.pic |
228 |
.SH AUTHOR |
229 |
Greg Ward |
230 |
.br |
231 |
Paul Heckbert supplied code for perspective projective mapping |
232 |
.SH "SEE ALSO" |
233 |
icalc(1), pcomb(1), pcond(1), pfilt(1), ximage(1) |