
Falcon 4M Camera Manual
Teledyne DALSA
03-032-20044-03
49
A Note on the Color Model and FFC
Flat field correction can be u sed in the color cam eras to both correct FPN and PRN U, as
on m ono cam eras, and p erform w hite balancing. H ow ever, w e recom m end that the u ser
u se the factory calibrated FFC coefficients for PRN U and FPN correction , and the d igital
system gain (
ssg
com m and ) to p erform w hite balancing.
For exam p le, w hen a m onochrom e sensor im ages a u niform w hite target illu m inated by a
halogen light sou rce, each p ixel ou tp u ts ap p roxim ately the sam e DN valu e. When the
sam e target is im aged by a color sensor, the red p ixels m ay p rod u ce m ore signal than the
green p ixels, w hich in tu rn p rod u ce m ore than the blu e p ixels.
White balancing involves increasing the gain of the blu e and green p ixels su ch that they
m atch the valu es from the red . When all color channels are m atched the im age w ill look
w hite w hen interp olated by the fram e grabber or host PC.
From the FFC algorithm‘s point of view this is simply a special case of PRN U w hich
resu lts in large d ifferences for each of the PRN U coefficients, on a p ixel -p er-p ixel basis, in
ord er to com p ensate for color.
The sam e p rincip les ap p ly to color as to m ono w ith the excep tion that you can exp ect to
see a w id er range of PRN U coefficient valu es.
H ere is an exam p le of a typ ical FFC calibration op eration u sing a halogen light sou rce:
1)
The cam era is p laced in
sem 2
(no other exp osu re m od e w ill allow FFC calibration).
2)
Settings su ch as fram e rate, exp osu re tim e, etc. are set as close as p ossible to the actu al
op erating cond itions. Set d igital gain to x1 (
ssg 0 4096
) and backgrou nd su btract
to 0 (
ssb 0 0
), as these are the d efau lts d u ring FFC calibration.
3)
Place the cam era in the d ark and send the ccf com m and , this fu nction p erform s the
FPN correction and au tom atically saves the FPN coefficients to non -volatile m em ory.
4)
Set
epc 1 0
, w hich enables the FPN correction and verify that the signal ou tp u t is
close to 0 DN . Leave
epc 1 0
for the next step as the
cpa
target assu m es that there
is no FPN . This is im p ortant on the 4M60/ 30 d u e to the large d ark offset valu es.
5)
Illu m inate the sensor, su ch that w ith
epc 1 0
, the red channel (i.e. red p ixels)
reaches 50-70 % satu ration. H ere w e assu m e that the red channel is the brightest and
the blu e channel the w eakest.
Send
cpa 2 T
w here T is typ ically 1.3x the average red channel ou tp u t level. This step is
im p ortant becau se if the target is too low (< 1.1x), then som e p ixels m ay not be able to
reach fu ll sw ing (1023 DN ) d u e to other corrections ap p lied by the cam era. Also ensu re
that the blu e channel is not too d im . The d efau lt sp m setting is 8x w hich w ill allow for a
m axim u m PRN U coefficient equ al to 8x; hence the red channel shou ld be LESS than 8x
brighter than the blu e (typ ically 4-5x is w hat w e see in the factory).
In the factory, for color cam eras only, w e u se a halogen light follow ed by a BG38 to act as
a light sou rce. The effective color tem p eratu re of this light is abou t 5200 K and its sp ectral
d istribu tion is show n in the figu re below .