on-orbit modulation transfer function (mtf) measurements ... · on-orbit modulation transfer...

On-orbit Modulation Transfer Function (MTF) Measurements for

IKONOS and QuickBird

Civil Commercial Imagery Evaluation Workshop

3/15/2006

Dennis Helder, Jason Choi, and Cody AndersonImage Processing LaboratoryElectrical Engineering and Computer Science DepartmentSouth Dakota State University

https://ntrs.nasa.gov/search.jsp?R=20070038252 2019-02-08T18:58:53+00:00Z

2

Outline• Introduction• Techniques• Target/Site Description• Results

– Ikonos– Quickbird

• Conclusions

Other IP Lab Contributors:Tim RugglesJim DewaldShriharsha MadhavanRuss VanDerWerffAmit AngalCam HelderLarry LeighDave Aaron

This work was supported by NASA grant NNS04AB66C

3

Introduction

• Point Spread Function– A method of evaluating the spatial

resolution of an imaging system.– A measure of the spread of a single

point of light. • Modulation Transfer function

(MTF)– MTF is a measure of the spatial

frequency response.– MTF is often calculated from the point

spread function (PSF).– System response at the Nyquist

frequency (or 0.5 cycle/pixel) is often used as a figure of merit.

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1MTF

Nor

mal

ized

val

ue

Normalized frequency

Value at Nyquist = 0.0568

4

• 2-dimensional PSF and MTF are difficult to obtain.• Often 1 dimensional functions are used:

– 1-D PSF is the line spread function (LSF).– LSF can be obtained by differentiation of the edge spread function

(ESF).

)},({),( yxPSFH yx ℑ=ωω

)0,0(),(

),(H

HMTF yx

yx

ωωωω =

5

Techniques• Edge Method

– Sub-pixel edge locations were found by Fermi function fit.– A least-square error line was calculated through the edge locations.– Modified Savitzky-Golay filtering was applied on each line.– The filtered profile was differentiated to obtain LSF– MTF calculated by applying Fourier transform to LSF.

Edge Method

6

• Pulse method– A pulse input is given to the imaging system.– Output of the system is the resulting image.– Edge detection and mSG filtering was applied to obtain

output profile. – Take Fourier transform of the input and output.– MTF is calculated by dividing output by input and

normalizing DC component to unity.

Pulse method 10 12 14 16 18 20 22 24

2

4

6

8

10

12

14

Edge detection

Pix

els

Pixels

Curve inflection pointLeast square fit line

7

SNR Definition

8

Target Description

Field Plan

9

Field campaign pictures on 6-22-2005

10

• NASA Stennis Tarp Target– Radiometrically and spectrally stable target with a large

DN difference from 3.6% and 52.1% reflectance panels.

– Edge oriented to obtained sub-sampled edge profile.– Blue tarps oriented at same angle.

NASA Stennis tarps

6°True North

Edge Angle

11

IKONOS Acquisitions• IKONOS Scene Information

Date Sensor Targets Resampling or MTF processing Product Type

CC / MTFC On Standard Geometrically Corrected

CC / MTFC Off Standard Geometrically Corrected

8/1/2005 IKONOSStennis tarps

Blue tarpsMirrors

12

Test image

Pix

el

Pixel

Angle=5.404 [deg]

5 10 15 20 25 30 35

5

10

15

Sub-pixel edge location

Least square error line

-5 -4 -3 -2 -1 0 1 2 3 4 5200

400

600

800

1000

1200

1400mSG filtering with raw data

DN

Pixel

DNdiff

= 1040.92STD

ave = 11.25

SNR = 92.54

Raw data

mSG filtering

-5 -4 -3 -2 -1 0 1 2 3 4 5-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2

FWHM = 1.6226 [Pixel]

Trimmed LSF

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Pixel0 0.1 0.2 0.3 0.4 0.5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2MTF


Nor

mal

ized

val

ue


Date Sensor Band Product Target Elevation Azimuth

8/1/2005 IKONOS Pan. CC/MTF off Stennis 66.0 118.7

13

Test image

Pix

el

Pixel

Angle=5.384 [deg]

5 10 15 20 25 30 35

5

10

15



-5 -4 -3 -2 -1 0 1 2 3 4 5

200

400

600

800

1000

1200

1400mSG filtering with raw data

DN

Pixel

DNdiff = 1043.96STDave = 12.93

SNR = 80.74

Raw data

mSG filtering

-5 -4 -3 -2 -1 0 1 2 3 4 5-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2


Trimmed LSF

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Pixel0 0.1 0.2 0.3 0.4 0.5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2MTF


Nor

mal

ized

val

ue



8/1/2005 IKONOS Pan. CC/MTF on Stennis 66.0 118.7

14

5 10 15

2

4

6

8

10

12

Pixel

Pix

el

Test image

Angle = 5.112 [deg]

Fermi edge location


-4 -3 -2 -1 0 1 2 3 4 5 6 7300

400

500

600

700

800

900

1000

1100Modified S-Golay interpolation with raw data

DN

Pixel

DNdiff = 629.36

STDave = 8.17

SNR = 77.03

Raw data

Modified S-Golay filtering

-3 -2 -1 0 1 2 3 4 5 6-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Pixel

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Trimmed LSF

FWHM = 3.0169 [Pixel] Output

Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50

60

Normalized Frequency

Fou

rier

tran

sfom

val

ue o

f in

put

and

outp

ut Fourier tranform of input and ouput

Discrete Output

Discrete Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50

60

70


Fou

rier

tran

sfor

m

Continuous Fourier tranform

Discrete Output

Continuous Input

0 0.1 0.2 0.3 0.4 0.50

0.10.20.30.40.50.60.70.80.9

11.11.2

MTF from Continuous Input



Nor

mal

ized

MT

F v

alue


8/1/2005 IKONOS Blue CC/MTF off Blue tarp 66.0 118.7

15

5 10 15

2

4

6

8

10

12

14

Pixel

Pix

el

Test image

Angle = 5.850 [deg]

Fermi edge location


-4 -3 -2 -1 0 1 2 3 4 5 6 7300

400

500

600

700

800

900

1000


DN

Pixel

DNdiff = 674.37

STDave = 8.43

SNR = 79.98

Raw data


-3 -2 -1 0 1 2 3 4 5 6-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Pixel

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Trimmed LSF


Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50

60


Fou

rier

tran

sfom

val

ue o

f in

put

and

outp


Discrete Output

Discrete Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50

60

70


Fou

rier

tran

sfor

m


Discrete Output

Continuous Input

0 0.1 0.2 0.3 0.4 0.50

0.10.20.30.40.50.60.70.80.9

11.11.2




Nor

mal

ized

MT

F v

alue


8/1/2005 IKONOS Blue CC/MTF on Blue tarp 66.0 118.7

16

Multi-year IKONOS Comparison• Consistent FWHM: 1.61 +/- 0.08• Values at Nyquist frequency were very stable.

– Mean = 0.11, STD = 0.01• No trends in PSF/MTF over 5 years.

-5 -4 -3 -2 -1 0 1 2 3 4 5-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Pixel

LSF

LSF over plot for Big Spring target (2000~2001) and Setnnis tarp (2002, 2005)

3-26-00 17:13 - FWHM= 1.7597 [pixel], SNR = 97.23-26-00 17:14 - FWHM= 1.6221 [pixel], SNR = 108.35-31-2001 - FWHM= 1.5826 [pixel], SNR = 45.76-22-2001 - FWHM= 1.5352 [pixel], SNR = 62.48-05-2001 - FWHM= 1.6380 [pixel], SNR = 79.79-26-2001 - FWHM= 1.4733 [pixel], SNR = 60.97-22-2002 - FWHM= 1.6175 [pixel], SNR = 113.68-01-2005 - FWHM= 1.6226 [pixel], SNR = 92.5

3-26-00 17:13

3-26-00 17:145-31-01

6-22-01

8-05-01

9-26-017-22-02

8-01-05

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.2

0.4

0.6

0.8

1

1.2

Pixel

MTF over plot for Big Spring target (2000~2001) and Setnnis tarp (2002, 2005)

3-26-00 17:13- at Nyquist = 0.09233-26-00 17:14- at Nyquist = 0.10545-31-01- at Nyquist = 0.10996-22-01- at Nyquist = 0.12078-05-01- at Nyquist = 0.10349-26-01- at Nyquist = 0.13757-22-02- at Nyquist = 0.11198-01-05- at Nyquist = 0.1117

MT

F

3-26-00 17:13

3-26-00 17:145-31-01

6-22-01

8-05-01

9-26-017-22-02

8-01-05

Date Sensor Band Product Target

8/1/05, 7/22/02, 9/26/01, 8/5/01, 6/22/01, 5/31/01, 3/26/00 IKONOS Pan. CC/MTFC

offStennis / Big

Springs

17

IKONOS SDSU campus on 8/1/2005CC Without MTFCCC With MTFC

Noisier & Sharper than CC

18

IKONOS Summary• Very consistent, high-quality sensor

– Pan FWHM = 1.61 +/- 0.08– Pan MTF @ Nyquist = 0.11 +/- 0.01– Multispectral MTF @ Nyquist = 0.52 (2005 data)

• ‘MTFC on’ processing provides increased MTF response with typical trade-off of increased contrast with some additional noise in Pan, less in Blue band.– Pan FWHM = 1.08 +/- 0.10– Pan MTF @ Nyquist = 0.48 +/- 0.08

• No indication of sensor degradation in five years.

19

Quickbird Acquisitions• Quickbird scene information

Date Sensor Targets Resampling or MTF processing Product Type

ResamplingKernel= CC

Standard2A(Radiometrically corrected)

MTF Standard2A

CC Standard2A

MTF Standard2A

CC Standard2A

MTF Standard2A10/18/2005 QuickBird

Stennis tarpsBlue tarps

Mirrors CC Standard2A

6/22/2005 QuickBirdVertical Stennis

tarpsBlue tarps

Mirrors MTF Standard2A

10/5/2004 QuickbirdBlue tarps

Mirrors

8/30/2004 QuickbirdBlue tarps

Mirrors

20

Quickbird 8-30-2004 Quickbird 10-5-2004

QB 6-22-2005 QB 10-18-2005

21

Test image

Pix

el

Pixel

Angle=-5.787 [deg]

10 20 30 40 50 60

5

10

15

20

25



-5 -4 -3 -2 -1 0 1 2 3 4 5

200

400

600

800

1000

1200

mSG filtering with raw data

DN

Pixel


SNR = 103.62

Raw data

mSG filtering

-5 -4 -3 -2 -1 0 1 2 3 4 5-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2


Trimmed LSF

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Pixel0 0.1 0.2 0.3 0.4 0.5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2MTF


Nor

mal

ized

val

ue


Date Sensor Band Resampling Target Elevation Azimuth

6/22/2005 Quickbird Pan. CC Stennis along 88.2 75.9

22

Test image

Pix

el

Pixel

Angle=-5.681 [deg]

10 20 30 40 50 60

5

10

15

20

25



-5 -4 -3 -2 -1 0 1 2 3 4 5

200

400

600

800

1000

1200

1400

mSG filtering with raw data

DN

Pixel


SNR = 65.97

Raw data

mSG filtering

-5 -4 -3 -2 -1 0 1 2 3 4 5-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2


Trimmed LSF

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Pixel0 0.1 0.2 0.3 0.4 0.5

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1

1.2MTF


Nor

mal

ized

val

ue


Along-track Stennis tarp target MTF result of QuickBird on 6/22/2005 (MTF)


6/22/2005 Quickbird Pan. MTF Stennis along 88.2 75.9

23

5 10 15

5

10

15

20

Pixel

Pix

el

Test image

Angle = 5.912 [deg]

Fermi edge location


-4 -3 -2 -1 0 1 2 3 4 5 6 7

200

400

600

800


DN

Pixel

DNdiff = 612.68

STDave = 5.41

SNR = 113.16

Raw data


-3 -2 -1 0 1 2 3 4 5 6-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Pixel

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Trimmed LSF


Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50


Fou

rier

tran

sfom

val

ue o

f in

put

and

outp


Discrete Output

Discrete Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50


Fou

rier

tran

sfor

m


Discrete Output

Continuous Input

0 0.1 0.2 0.3 0.4 0.50

0.10.20.30.40.50.60.70.80.9

11.11.2




Nor

mal

ized

MT

F v

alue


6/22/2005 Quickbird Blue CC Blue tarp cross 88.2 75.9

24

5 10 15

5

10

15

20

Pixel

Pix

el

Test image

Angle = 5.699 [deg]

Fermi edge location


-4 -3 -2 -1 0 1 2 3 4 5 6 7

200

400

600

800


DN

Pixel

DNdiff = 761.86

STDave = 6.75

SNR = 112.84

Raw data


-3 -2 -1 0 1 2 3 4 5 6-0.2

0

0.2

0.4

0.6

0.8

1

1.2

Pixel

Nor

mal

ized

Diff

eren

tiatio

n va

lue

Trimmed LSF


Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50


Fou

rier

tran

sfom

val

ue o

f in

put

and

outp


Discrete Output

Discrete Input

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

10

20

30

40

50


Fou

rier

tran

sfor

m


Discrete Output

Continuous Input

0 0.1 0.2 0.3 0.4 0.50

0.10.20.30.40.50.60.70.80.9

11.11.2




Nor

mal

ized

MT

F v

alue


6/22/2005 Quickbird Blue MTF Blue tarp cross 88.2 75.9

250 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

0

0.2

0.4

0.6

0.8

1

1.2

Normalized frequency [cycle/pixel]

Nor

mal

ized

MT

F

Stennis tarp MTF plots on 6/22/05 (CC) 10/18/05 (MTF) from QuickBird

6/22/05, CC = 0.0364 6/22/05, MTF = 0.104310/18/05, CC = 0.093810/18/05, MTF = 0.2581

6/22/05,CC, along-T.

6/22/05, MTF, along-T.

10/18/05, CC, across-T.

10/18/05, MTF, across-T.

QuickBird panchromatic band Along / Cross Track Direction Comparison

• First estimate of along-track PSF/MTF.• Along-track PSF/MTF not significantly different than cross-

track.

-5 -4 -3 -2 -1 0 1 2 3 4 5-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

Pixel

Nor

mal

ized

LS

F

Stennis tarp LSF plots on 6/22/05 (CC) 10/18/05 (MTF) from QuickBird

6/22/05 CC = 1.8939 [pixel], SNR = 103.6 6/22/05 MTF = 1.6925 [pixel], SNR = 66.010/18/05 CC = 1.8945 [pixel], SNR = 44.210/18/05 MTF = 1.7109 [pixel], SNR = 56.4

6/22/05, CC, along-T.

6/22/05, MTF, along-T.

10/18/05, CC, across-T.

10/18/05, MTF, across-T.

Date Sensor Band Resampling Target

6/22/05, 10/18/05 Quickbird Pan. MTF / CC Stennis

26-3 -2 -1 0 1 2 3 4 5 6 7

-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Pixel

Nor

mal

ized

Pul

se R

espo

nse

Fun

ctio

n (P

RF

)

PRF plots for QuickBird on 6/22/05 and 10/18/05 with CC interpolated scenes

6/22/05 Cross-T. FWHM = 2.3216 [pixel], SNR = 113.210/18/05 Cross-T. FWHM= 2.3264 [pixel], SNR = 94.6 6/22/05 Along-T. FWHM = 2.4979 [pixel], SNR = 72.010/18/05 Along-T. FWHM= 2.4777 [pixel], SNR = 22.4

6/22/05 Cross-T.

10/18/05 Cross-T.6/22/05 Along-T.

10/18/05 Along-T.

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.2

0.4

0.6

0.8

1

1.2


MTF plots for QuickBird on 6/22/05 and 10/18/05 with CC interpolated scenes

6/22/05 Cross-T. MTF=0.167910/18/05 Cross-T. MTF=0.1420 6/22/05 Along-T. MTF=0.127010/18/05 Along-T. MTF=0.1095

Nor

mal

ized

MT

F

6/22/05 Cross-T.

10/18/05 Cross-T.

6/22/05 Along-T.

10/18/05 Along-T.

QuickBird Blue Band Along / Cross Track Direction Comparison

• Cross-track profile exhibits narrower PRF FWHM and under shoots. • System MTF shape differs in the orthogonal directions.


6/22/05, 10/18/05 Quickbird Blue CC only

Blue tarps Cross / along-track

27

QuickBird Blue Band Along / Cross Track Direction Comparison

• FWHM values were reduced by MTF resampling process.• MTF values were increased—most significantly in cross track

direction.

-3 -2 -1 0 1 2 3 4 5 6 7-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

Pixel

Nor

mal

ized

Pul

se R

espo

nse

Fun

ctio

n (P

RF

)

PRF plots for QuickBird on 6/22/05 and 10/18/05 with MTF resampled scenes

6/22/05 Cross-T. FWHM = 2.0699 [pixel], SNR = 112.810/18/05 Cross-T. FWHM= 2.0988 [pixel], SNR = 86.1 6/22/05 Along-T. FWHM = 2.2864 [pixel], SNR = 33.310/18/05 Along-T. FWHM= 2.2398 [pixel], SNR = 44.4

6/22/05 Cross-T.

10/18/05 Cross-T.

6/22/05 Along-T.

10/18/05 Along-T.

0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.50

0.2

0.4

0.6

0.8

1

1.2


Nor

mal

ized

MT

F

MTF plots for QuickBird on 6/22/05 and 10/18/05 with MTF resampled scenes

6/22/05 Cross-T. MTF=0.230610/18/05 Cross-T. MTF=0.2070 6/22/05 Along-T. MTF=0.272910/18/05 Along-T. MTF=0.2153

6/22/05 Cross-T.

10/18/05 Cross-T.

6/22/05 Along-T.

10/18/05 Along-T.


6/22/05, 10/18/05 Quickbird Blue MTF only

Blue tarps Cross / along-track

28

Band Resamp. mean sigma mean sigma Obs.

Pan. CC Cross 1.89 0.09 1Along 1.89 0.04 1

MTF Cross 1.71 0.26 1Along 1.69 0.10 1

Blue CC Cross 2.32 0.01 0.18 0.07 4Along 2.49 0.01 0.12 0.01 2

MTF Cross 2.11 0.05 0.30 0.17 4Along 2.27 0.02 0.25 0.03 2

Quickbird Results Summary 2004-2005Scan

Direction

FWHM (gsd)

MTF at Nyquist (cycles/gsd)

Notes:1. Blue Band FWHM’s are for Pulse Response Functions.2. Only one Pan band observation! Essentially same PSF/MTF in

both scan directions.3. MTF resampling provides noticeable contrast improvement;

SNR is lowered, but still acceptable.4. Overall SNR is good to excellent.5. PRF is very repeatable in Blue band; MTF is noticeably less so

29

Quickbird SDSU campus on 6/22/2005CC resamplingMTF resampling MTF resampling CC resampling

Noisier & Sharper than CC

30

2002/2003/2005 QuickBird Comparisons

• GSD changed from 0.7 to 0.6 meters after 2003.

• Spatial resolution performance appears consistent from 2002 to 2005.

-3 -2 -1 0 1 2 3-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

meters

Nor

mal

ized

LS

F

LSF over plot for Stennis tarp

7/20/02, CC, FWHM= 1.0755 [pixel], SNR = 100.1

8/25/02, CC, FWHM= 1.0436 [pixel], SNR = 100.5

9/7/02, CC, FWHM= 1.0228 [pixel], SNR = 141.39/15/03, CC, FWHM= 0.8596 [pixel], SNR = 110.4

10/18/05, CC, FWHM= 1.1367 [pixel], SNR = 44.2

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 10

0.2

0.4

0.6

0.8

1

1.2

Normalized frequency [cycle/meter]

Nor

mal

ized

MT

F

MTF over plot for Stennis tarp target

7/20/02, CC, MTF = 0.15248/25/02, CC, MTF = 0.1575

9/7/02, CC, MTF = 0.1786

9/15/03, CC, MTF = 0.1449

10/18/05, CC, MTF = 0.0938

Nyquist location in 2002Nyquist location after 2003


7/20/02, 8/25/02, 9/7/02, 9/15/03, 10/18/05 Quickbird Pan. CC Stennis tarp

31

Overall Quickbird Conclusions

• Initial along-scan PSF/MTF estimates indicate slightly more blur than cross-scan, as expected.

• Good to excellent SNR.• MTF compensation provides noticeable contrast

boost with normal loss of SNR in Panchromatic band, but minimal loss of SNR in Blue band.

• No degradation of Panchromatic band PSF/MTF indicated from 2002 through 2005.

on-orbit modulation transfer function (mtf) measurements ... · on-orbit modulation transfer...

Documents