and
zed
ain,
tor.
ese
ved
(3)
(4)
ble
vith
ixel
low
ark
ere
1ed
69
during DC restore. DC restore is a process in which the output is set equal to a non-zero constant number of DN (in this case,
two). This very process invalidates any data taken during that time which is meant to characterize the detector response. The
data from window w2 in forward scans was positioned in the DC restore region. Ittherefore failed to correct for all of the scan to
scan variations. A comparison of dark shutter responses from Landsat 5, using both the window in the Long Term Parameter
File and the new window w6 used in this study, can be found in Figure 2.
(a) Sample forward scan (b) Sample reverse scan
250 "T T T T T 250 [^ T
- Lamp pulse : t Lamp pulse
200 = À 2200 = du
= [ a [^
a, sor - = 150 LA
al " : d M.
tC. oo E 100
+— N, e
50 E w2 w1 (nf so L- I
E pd t w4 - WE Jg 7
oo da — DC Bestora W3 \ t i-e iasiore 1
o 200 400 600 800 o 200 age . 800 800
Relative Minor Frame Relative Minor Frame
Figure 1. Various window positions during dark shutter collection attempted during this study.
(a) Using window from Landsat 5
: Equalization source
Long Term Parameter File
225 EY Ed
—- Nominal (no equaliz) '
— EDC118 scene — 1984
--- Libya scene — 1991 |
ak T^
220
7
; a
[ 7 <
[- 3 E
> ; =
B 1
Le F 3 E
o - —
SB a, mh a
2
7 | z :
= : 3 8 aol zd
sr = = [ 7
o F Scon Direction Code: j "M 1
» [ a © L
< EE = = ee = = se se se ss se se GO cs i 1 is
E RUNI TEC EET UP m mem TP mem = 10 55
ter lee C RANG Re 77 $00 Detector ID
Line no. (along-track direction) Figure 3. Detector Gains Determined from Equalization
; : of 1984 and 1991 Scenes vs. Nominal Gains
Y (b) Using Wirdow we : . from Long Term Parameter File (Band 1)
or Af:er rad. co
z 3 15 bit
a r 8 bit
am o 3- HE
© e £ ie A
£^ S MAT SAE TN
7) 9 WM M gar
© 1 S
$ N
<< t REV scon
oO A L 1 1 A 5 2r €000000000000000 0000000000000000 =
o 100 200 300 400 500 a L
Line no. (along-track direction) [ F2 scan
Q A | du 4
18C 200 220 240
Figure 2. Average Dark Shutter Response for Each Line Line no. (along-track direction)
Using Different Shutter Extraction Windows
Sample Data Taken from Libyan Desert 1991 Scene Band 4
(Note: Lower valued areas in Figure (b) are reverse scan)
Figure 4. Along Track Profiles Through Uniform Lake
Quantized to 8 and 15 Bit Precision Levels
Later in the study, some imagery processed with this technique began to show residual striping that varied in intensity with
scan position, in this case peaking at the right side of the image. This is a phenomenon sometimes called scan droop. A
comparison of average shutter values before and after DC restore indicated that the detector response apparently gets
"pumped up" after DC restore and slowly drops a fraction of a DN during the scan. A new technique was developed based
upon the assumption that the DC restore in each scan was reinitializing the detector response to its higher value. Two
average samples were collected during each dark shutter using windows w6 and w7. Then, using the average calculated after
IAPRS, Vol. 30, Part 5W1, ISPRS Intercommission Workshop "From Pixels to Sequences”, Zurich, March 22-24 1995