subjective image quality metrics from the wave...
TRANSCRIPT
Subjective Image Quality Metrics fromThe Wave Aberration
David R. WilliamsWilliam G. Allyn Professor of Medical Optics
Center For Visual ScienceUniversity of Rochester
Commercial Relationship:Bausch and Lomb
Funding:Bausch and Lomb
NSF Science and TechnologyCenter for Adaptive Optics
Collaborators:Univ of Rochester:Li ChenNathan DobleHeidi Hofer
Ian Cox, Bausch and LombRay Applegate, University of HoustonLarry Thibos, Indiana University
Scott MacRaeJason PorterBen SingerRemy Tumbar
How Bad Is This Wave Aberration?
Goal:
To compute a number that captures the subjective effect of the eye’s wave aberration.
Uses:
Assessing severity of the wave aberrationCalculating the best correction
How Bad Is This Wave Aberration?
RMS Wavefront Error = 0.87 µm
Some Aberrations Interact Strongly in Image Blur
Spherical Aberration
rms = 0.16 µm
Defocus andSpherical Aberration
rms = 0.52 µmDefocus
rms = 0.5 µm
Zernike Modes
trefoil coma coma trefoil
quadrafoil secondary spherical secondary quadrafoil
secondary secondarysecondary secondary
defocusastigmatism astigmatism
astigmatism astigmatism
trefoiltrefoil comacoma
Z20Z 2
-2 Z22
Z 3-1 Z3
1Z 3-3 Z3
3
Z40 Z4
2Z 4-2 Z4
4Z 4-4
Z51Z 5
-1 Z53Z 5
-3 Z55Z 5
-5
Higher OrderAberrations
2nd
radialorder
Lower OrderAberrations
3rd
4th
5th
pentafoilpentafoil
Zernike Modes
trefoil coma coma trefoil
quadrafoil secondary spherical secondary quadrafoil
secondary secondarysecondary secondary
defocusastigmatism astigmatism
astigmatism astigmatism
trefoiltrefoil comacoma
Z20Z 2
-2 Z22
Z 3-1 Z3
1Z 3-3 Z3
3
Z40 Z4
2Z 4-2 Z4
4Z 4-4
Z51Z 5
-1 Z53Z 5
-3 Z55Z 5
-5
Higher OrderAberrations
2nd
radialorder
Lower OrderAberrations
3rd
4th
5th
pentafoilpentafoil
Zernike Modes
trefoil coma coma trefoil
quadrafoil secondary spherical secondary quadrafoil
secondary secondarysecondary secondary
defocusastigmatism astigmatism
astigmatism astigmatism
trefoiltrefoil comacoma
Z20Z 2
-2 Z22
Z 3-1 Z3
1Z 3-3 Z3
3
Z40 Z4
2Z 4-2 Z4
4Z 4-4
Z51Z 5
-1 Z53Z 5
-3 Z55Z 5
-5
Higher OrderAberrations
2nd
radialorder
Lower Order Aberrations
3rd
4th
5th
pentafoilpentafoil
Zernike Modes
trefoil coma coma trefoil
quadrafoil secondary spherical secondary quadrafoil
secondary secondarysecondary secondary
defocusastigmatism astigmatism
astigmatism astigmatism
trefoiltrefoil comacoma
Z20Z 2
-2 Z22
Z 3-1 Z3
1Z 3-3 Z3
3
Z40 Z4
2Z 4-2 Z4
4Z 4-4
Z51Z 5
-1 Z53Z 5
-3 Z55Z 5
-5
Higher OrderAberrations
2nd
radialorder
Lower OrderAberrations
3rd
4th
5th
pentafoilpentafoil
There are strong interactions between Zernike Modes.
Therefore,Decomposing the wave aberration
into Zernike modes is not the best way to evaluate the subjective impact of
the wave aberration
How Bad is This Wave Aberration?
Point Spread FunctionWave Aberration
Use Retinal Image Quality, Not the Wave Aberration
Aberrations in Lens and Cornea Distort Wave front
Principle of Adaptive Optics
Deformable MirrorCorrects Wave front
Sharp Image in Camera
Wave front SensorMeasures Wave front
Adaptive Optics Sharpens the Eye’s Point Spread Function
QuickTime™ and aGraphics decompressor
are needed to see this picture.
Adaptive Optics Can Create Wave Aberrations (Subject: ND�)
Wave Aberration After AO correction With coma added
Convoluted retinal image
Wave aberrations from Lasik postOp patient
Wave aberrations created by adaptive optics (With real eye, JP)
Blur Matching
• Lots of Sharp Edges• Edges At All Orientations• Seen Through Adaptive
Optics
Binary Noise Stimulus
550 nm, 1 deg, 6 mm pupil
The subject adjusted the amplitude of defocus so that thesubjective blur matched that of the patient wave aberration.
Blur Matching of Patient Wave Aberrations
-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20
Patient wave aberration Stimulus
QuickTime™ and aPhoto - JPEG decompressor
are needed to see this picture.
Zernike mode Zernike mode
Am
plitu
de(µ
m)
Def
ocus
(D)
Defocus
Blur Matching Controls for Neural Differences between
Patients
Using Multiple Subjects Controls for Neural Adaptation
Acuity does not always capture the subjective quality of vision
Equal Acuity But Different Subjective Image Quality
Compare Subject Matches with Matches Made Using Three Different Metrics
Wavefront RMSStrehl Ratio
Neural Sharpness
Wave Aberration RMS
Big RMS Small RMS
-3 -2 -1 0 1 2 3-1
0
1
2
3
Am
plitu
de (µ
m)
Aperture(mm)
Big RMS
Small RMS
R2 = 0.4627
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1 1.2
RMS Metric
Met
ric
mat
ches
(D)
Subject Matches (D)
Strehl Ratio of Point Spread Function
Diffraction-limited PSF(Perfect Eye)
Heye
Actual PSF(Aberrated Eye)
Strehl Ratio = H eye
H dl
C. of Austin Roorda
Hdl
Strehl Ratio Metric
R2 = 0.4902
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1 1.2
Met
ric
mat
ches
(D)
Subject Matches (D)
A Simple, Biologically-Plausible Metricfor Subjective Image Quality
( ) SubjectiveImage
Quality Point Spread
FunctionGaussian Neural Blur
σ = 0.8’
Σ x =
Neural Sharpness Metric
R2 = 0.703
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1 1.2
Met
ric
mat
ches
(D)
Experiment (D)
Wavefront RMS Neural Sharpness
R2 = 0.703
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1 1.2
Met
ric
mat
ches
(D)R2 = 0.4627
0
0.2
0.4
0.6
0.8
1
1.2
0 0.2 0.4 0.6 0.8 1 1.2
Met
ric
Mat
ches
(D)
Subject Matches (D)
Collaboration to Identify the Optimum Image Quality Metric
Ray Applegate, University of Houston:Effectiveness of Image Quality Metrics in PredictingVisual Acuity with Convolution Simulations
David Williams, University of RochesterEffectiveness of Image Metrics in predictingSubjective Image Quality with Adaptive Optics
Larry Thibos, Indiana University:Effectiveness of Image Quality Metricsin Predicting Visual Acuity in the Population
defocus
met
ric v
alue
maximum
Optimizing refraction with an image quality metric
error
subjective
search in 3-D space
Guirao and Williams (2003)
Conclusions• Generating blur with adaptive optics leads
to a robust metric for correcting vision.
• It is hard to estimate subjective blur from the wave aberration.Zernike Decomposition doesn’t helpmuch.
• The point spread function combined with a biologicallyplausible model of neural bluris better.
• Standards for optimizing correction from wavefrontare in the works