isscc iso baer - kilopixelkilopixel.net/publications/isscc_iso_baer.pdf · 2013. 8. 4. · iso...
Post on 18-Aug-2020
2 Views
Preview:
TRANSCRIPT
Rick Baer
February 10, 2005
ISO Speed
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 2
Outline
IntroductionUnderstanding ISO speedISO speed characterizationISO speed and image processingSummary
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 3
Introduction•ISO speed is useful in still photography because it determines the nominal exposure conditions.•The ISO (ASA) speed metric was originally developed to describe the sensitivity of silver-halide film. The relationship between speed and image quality is only implicit.•The ISO 12232 standard defines an ISO speed metric for digital cameras (and solid-state image sensors) that is explicitly related to image quality. •ISO speed does not apply to imaging at non-visible wavelengths.
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 4
Understanding ISO speed
Exposure controls & meteringISO speed of filmISO speed of a solid-state image sensor
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 5
Basic exposure controlsExposure ~ number of incident photons
aperture (f#) n.d. filter
(density)
shutter (period)
sensor
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 6
Why control exposure?
Dynamic rangeShutter speed controlAperture control – depth of field
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 7
Underexposure
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 8
Correct exposure
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 9
Overexposure
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 10
Slow shutter
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 11
Fast shutter
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 12
Wide aperture
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 13
Narrow aperture
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 14
Exposure index
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 15
How exposure is controlled with a lightmeter:
1. Set the exposure index (nominally equal to ISO speed)2. Select an aperture (f#) and read the shutter speed
-or-2. Select a shutter speed and read the aperture
4.15 EI)#( 2 La
tf
=
sensor/film exposure
index
average scene
luminanceshutter speed
aperture value
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 16
What is the exposure index?
Where: <Hg> = geometric mean focal plane exposure<Ha> = arithmetic mean focal plane exposure
ag HH108EI ≈=
[lux-seconds]
used in electronic imaging systems,
ISO 12232
Ef (focal plane illuminance) [lux]
<Ha> = <Ef> ⋅ Te
lens sensor
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 17
Relationship between scene illuminance and focal plane illuminance
( ) ( )22s
fm1#4
EE+
=f
R
focal plane illuminance
(lux)
scene illuminance
(lux)scene
reflectivity (lux)
image magnification
Typical scene illuminance levels:Direct sunlight 100,000 luxIndirect sunlight 10,000Overcast 1,000Office 300Pub 10Full moon 0.1
006.EE
s
f ≈
For R = 18%, f# = 2.8, m = 0(1.8 focal-plane lux with office lighting)
EsEf
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 18
Photopic units
( ) λλλ dy ev ∫∞
Φ=Φ0
680
λy
luminous flux
(lumens)
spectral flux
(Watts)
photopic response function
photopic units (lux) describe light intensity as perceived by a
human observer
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 19
How many photons are there?
1 lux = 1 lumen / m2 ~ 10000 photons/sec/um2
(for a spectrally broad illuminant)
Example: office lighting, 10 um2 pixel, 1/120 second exposure1500 photons/pixel
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 20
ISO speed and exposure index
•The correct exposure for a particular scene determines the exposure index.
•The ISO speed is equal to the exposure index for a statistically average scene.
exposure compensation
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 21
High-key scene
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 22
Low-key scene
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 23
Relationship between ISO speed and (digital) image quality
•The gain applied to a solid-state sensor can be adjusted to change the ISO speed•Image quality (SNR) depends on ISO speed
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 24
ISO = 200
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 25
ISO = 3200
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 26
ISO speed comparisons are meaningless unless image quality
is considered!
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 27
ISO speed of film
Speed range of commercial filmSpeed versus quality tradeoffISO measurement procedure
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 28
Color negative film
ISO 100 ISO 1600
B&W film
ISO 400
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 29
Speed versus quality
Film response is nonlinearFilm is a threshold detectorGrain noise is more important than shot noiseGrain size increases with film speed
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 30
Scanned film
(ISO 100) (ISO 800)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 31
Measurement apparatus
Calibrated light source
variable aperture
IR filter
integrating sphere
calibrated lux meter
camera body (no lens)
shutter speed control
processingdensitometry
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 32
Measurement procedure
1. Use aperture and shutter to vary focal plane exposure from frame to frame.
2. Process film for gamma = 0.6153. Measure film density versus focal plane exposure4. Plot density versus log exposure5. Determine exposure intercept at (fog + 0.1) density6. ISO speed = 0.8 / En
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 33
(film density plot)
ISO = 0.8/En = 0.8/10-2.2 = 125
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 34
The ISO characterization methods used for film are not applicable to solid-state image sensors, with linear responses
and different noise mechanisms.
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 35
ISO speed of a solid-state image sensor
Comparison of film and solid-state sensorsISO 12232 methodologyMonochrome image sensor modelColor image sensor model
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 36
Comparison of film and solid-state image sensors
camera
camera
film film developer
image sensor
image processor
may be integrated!
may incorporate camera functions (e.g. metering, shutter)
Image sensors and digital processing compare with film and developing
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 37
ISO 12232 methodology
•Determine focal plane exposure (H) required to obtain a particular SNR value•Specify ISO speed range:
• Snoise10 = 10/H @ SNR = 10 (first acceptable image) • Snoise40 = 10/H @ SNR = 40 (first excellent image)
•Specify ISO speed at saturation• Ssat = 78 / H @ saturation
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 38
H V
vary illumination, shutter speed to control focal plane
exposure
measure SNR of output signal
sensortest camera
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 39
H
SNR
10
20
30
40
HsatH10 H40
S = 10/H S = 78/H
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 40
ISO speed model for monochrome image sensors
•Determine H required to achieve specified SNR•ISO speed = 10/ H
x
QE
signal electrons
noise electrons
read noise, dark current
÷ SNR
x H
photopic response
noise model
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 41
Photopic quantum efficiency (photopic electrons per lux)
signal electrons
pixel array
focal plane lux
IR filterPlanckian radiator
photopic response function
( ) λλλ dy e∫∞
Φ=0
680H
( ) ( ) 1/5 1 −
−=Φ λ
λαλ kThc
e e
( ) λλλλλ dch
e∫∞ Φ
=0
e)(QE)(IRN
H / N ep =η
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 42
Typical photopic QE for a monochrome image sensor(Panasonic MN3776; peak QE = 50%)
Photopic QE
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
color temperature [K]
kilo
elec
tron
s/(lu
x-µ
m2 )
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 43
Noise model
signal electrons
[e-]
r.m.s. noise
electrons [e-]
read noise [e-]
= shot noise power Σ
xdark current [e-/sec]
integration time [sec]
X2
X2
sum of powers of uncorrelated noise
sources
x X2
PRNU
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 44
Derivation of ISO speed equation
2NHA
HANS
rp
p
x +=
η
η
Photopic QE pixel area
focal plane exposure
read noise
shot noise
1
2
2
2 )N/S(411
)N/S(A20
−
++=
xx
px
NrSη
•ISO noise speed increases linearly with QE and pixel area
•“Acceptable” noise speed (SNR=10) depends on electronic noise
(*** neglecting PRNU ***)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 45
SNR curvesSNR vs. charge
0
0.5
1
1.5
2
2.5
3
0 1 2 3 4 5 6
log10( electrons)
log1
0( S
NR
)
shot noise onlyNR = 10, PRNU = 0.01Nr = 10, PRNU = 0
Achievable SNR limited by PRNU
Read noise only affects “acceptable” ISO speed
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 46
ISO speed curves for a typical sensor
ηP = 3.7 ke, Nr =10 e, Nsat = 20 ke
Monochrome ISO speed
0
1000
2000
3000
4000
5000
2 3 4 5 6 7 8 9 10
pixel pitch [um]
ISO
spe
ed acceptableexcellentsaturation
(*** neglecting PRNU ***)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 47
Effect of IR and UV (wavelengths outside photopic response range)
Black body emission
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
2000 3000 4000 5000 6000 7000 8000 9000 10000
Temperature [K]
Rel
ativ
e ph
oton
flux
far IRnear IRvisibleUV
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 48
ISO speed ratio for BB source, w & w/o IR filter
Photopic response with IRF / photopic response w/o IRF
0
0.5
1
1.5
2
2.5
3
2500 3000 3500 4000 4500 5000 5500 6000 6500 7000
Color temperature [K]
Rat
io
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 49
Effect of dark current
TeNdA10 p
a
η≈S
In the low-SNR, high noise limit:
ISO “acceptable”
speeddark
current [e/sec]
exposure period
ISO speed varies inversely with exposure period: longer integration times don’t help!
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 50
The high ISO speeds of solid-state image sensors (compared to film) are a consequence
of the higher QE of silicon.
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 51
ISO speed characterization
Camera characterization: ISO 12232Sensor characterization: QE and noise
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 52
ISO 12232 procedures for determining speed
Direct focal plane exposure methodIndirect scene luminance method
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 53
Measurement apparatus (direct focal plane exposure)
Calibrated light source (fixed color
temperature)
variable aperture
IR (& UV) filter
integrating sphere
calibrated lux meter
camera body (no lens)
shutter speed control
raw digital output
baffle
Measure SNR versus focal plane exposure
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 54
Illumination control
d
z
f# = d/z
Sony ICX252 angular response
0%
20%
40%
60%
80%
100%
120%
0.00 5.00 10.00 15.00 20.00 25.00 30.00
Incidence Angle [deg.]
Res
pons
e [r
el.]
horzvert
f1.4f2.0f2.8F3.5 f0.6
without baffle
The angular spread of illumination must match the lens f#.
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 55
Other requirements
•Linear data (raw data or linearized)•No lossy compression•Proper white balance•Integration period < 1/30 s (may need aperture control)•Consider only central pixels when sensors include microlenses
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 56
SNR curve (monochrome CCD)
SNR vs. focal plane exposure
0
10
20
30
40
50
60
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Focal plane exposure (lux-sec)
SNR
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 57
ISO speed from SNR curve
SNR vs. focal plane exposure
0
10
20
30
40
50
60
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08
Focal plane exposure (lux-sec)
SNR
S10 = 10/0.0044 = 2300 S40 = 10/0.046 = 218
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 58
Measurement apparatus (indirect scene luminance: reflection)
lux metercamerashutter speed control
raw digital output
lens
Measure SNR versus (scene luminance ⋅ exposure time)
color meter
baffle
light source
variac
luminance meter
(alternative to luminance meter)
diffuse reflection
target
scene luminance
(used to adjust source color temperature)
(used to adjust luminance)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 59
Measurement apparatus (indirect scene luminance: transmission)
camerashutter speed control
raw digital output
lens
color meter
light source
variac
luminance meter
transmission target
scene luminance
diffuserbaffle
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 60
Equation relating focal plane exposure to target luminance
22a
a )1()#(10065
mftLH+
=
luminance Integration time
Calculate speed as 10/Ha @ SNR = x (as before)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 61
Need for OECF characterization: perform analysis on OECF target
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 62
SNR curve from OECF target
OECF target method
0.0
5.0
10.0
15.0
20.0
25.0
30.0
35.0
40.0
45.0
50.0
0 2 4 6 8 10 12
patch number
SNR
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 63
Sensor characterization: QE and noise
QE measurementNoise characterization
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 64
ISO speed calculation from QE and noise (monochrome case)
photopic QE
calculation
noise model
operating parameters (gain, temp.,
etc.)
QE(λ)
IR(λ)
noise parameters
ISO speed
equation
ISO speed range
spectral measurements
noise measurements
ηp
Nr
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 65
ISO speed calculation from QE and noise (color case)
photopic QE
calculation
noise model
operating parameters (gain, temp.,
etc.)
QE(λ)
IR(λ)
noise parameters
ISO speed model
ISO speed range
spectral measurements
noise measurements
ηp
Nr
color correction matrix
calculation
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 66
QE measurement
light source
variable aperture
low-pass filter
integrating sphere
reference photodiode
camera body (no lens)
shutter speed control
raw digital output
baffle
Measure QE versus λ
monochrometercalibration photodiode
reference output
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 67
CC matrix determination
From QE curveFrom Macbeth chart
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 68
Noise measurement
(see ISO 15739 standard)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 69
Example: CCD for a typical consumer digital camera
0%
10%
20%
30%
40%
50%
360 410 460 510 560 610 660 710 760
Wavelength [nm]
QE
[%]
Sony ICX202
Nr = 8 e
Nsat = 10 ke
PRNU = 1%
SA = 285
For 3 um pixel pitch:
SE = 23
Ssat = 72
Acceptable quality speed: Excellent quality speed:
Saturation speed:Speed range: ~ 50 - 400
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 70
Effect of signal processing on ISO speed
•Five (coupled) dimensions of image quality•Effects of common image processing functions
• Demosaic• Vignetting correction• Sharpening• Tone mapping• Color space conversion• Compression
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 71
Five (coupled) dimensions of image quality
The five “R’s” of image qualityResolution (pixel count, MTF)
snR (ISO noise speed)
dnR (ISO saturation speed)
Reproduction (exposure control, color and tone reproduction)
aRtifacts (demosaic and sampling artifacts, flare)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 72
Coupling between image quality dimensions
(high resolution, high-noise) (low resolution, low-noise)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 73
Coupling between ISO speed and other IQ dimensions
SNR (ISO speed)
DNR
ResolutionReproduction
aRtifacts
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 74
Coupling between ISO speed and other IQ dimensions
SNR (ISO speed)
DNR
ResolutionReproduction
aRtifacts
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 75
DemosaicBayer mosaic pattern
bilinear reconstruction
RR 8/5 σσ =′ BB 8/5 σσ =′GG 16/9 σσ =′
σRσ’R
Bilinear reconstruction reduces noise (at the expense of resolution)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 76
Vignetting correction
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 77
Sharpening
~ 30% noise increase
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 78
Tone mapping
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 79
Color space conversion
RGB 888 YUV 422, YUV 411 …
Chrominance down-sampling reduces color noise
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 80
Compression
original image JPEG compressed
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 81
SummaryISO speed is a valuable quality metric for the solid-state sensors used in (visible) photography.Different combinations of measurement and analysis can be used to obtain the ISO speed.Digital processing can improve one dimension of image quality at the expense of others. All dimensions of IQ must be considered together in sensor comparisons.Total IQ: (ISO speed) × resolution {also true for DNR}
∝ QE × (sensor area) Size matters!
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 82
Acknowledgements•Albert Theuwissen and the organizers of this forum •Jack Holm
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 83
References1. ISO 12232: Photography – Electronic still-picture cameras – Determination of ISO speed (1998)2. M. Kriss, “A model for equivalent ISO CCD camera speeds”, SPIE Vol. 3302, pg. 56-67 (1998)3. J. Holm, “The photographic sensitivity of electronic still cameras”, J. Soc. Photogr. Sci. Tech. Japan, Vol. 59, No. 1, pg.
117-131 (1996).4. R. Baer & J. Holm, “A model for calculating the potential ISO speeds of digital still cameras based upon CCD
characteristics”, IS&T PICS Conference, pg. 35-38 (1999). 5. R. Palum, “How many photons are there?”, IS&T PICS Conference, pg. 203-206 (2002).6. J. Holm, “Challenges and progress in digital photography standards”, SPIE Vol. 5294 (2004).
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 84
Appendices•ISO speed model for color image sensors
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 85
ISO speed model for color image sensors
Assumptions:1. Independent RGB color samples at every pixel location2. No image processing other than white balance and color
correction (e.g. no tone correction, sharpening, compression)
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 86
Extension of monochrome model to color
x
QE vector
R,G,B signal electrons
R,G,B noise electrons
read noise, dark current
÷ SNR
x H
photopic response
noise model
color correction
white balance RGB to Y
color correction
RGB to Y-cb-cr
Y-cb-cr to perceived
noise
white balance
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 87
Spectral response => photopic QE vector
( ) λλλλλ dch
e∫∞ Φ
=0
e)(QE)(IRN
H / N ep =η
SONY ICX224AQ
0%
10%
20%
30%
40%
360 460 560 660 760 860
Wavelength [nm]
QE
[%]
IR filter
0.0
0.2
0.4
0.6
0.8
1.0
360 460 560 660 760 860
Wavelength [nm]
Tran
smis
sion
5500K black body
0%
20%
40%
60%
80%
100%
360 460 560 660 760 860
Wavelength [nm]
Rel
ativ
e po
wer
[electrons/lux-second]
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 88
White balance
BGR
G/B0001000G/R
⋅=BGR
Calculate coefficients from photopic QE vector
Apply to noise vector
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 89
Color correction and spectral responseSONY ICX224AQ
0%
10%
20%
30%
40%
360 460 560 660 760 860
Wavelength [nm]
QE
[%]
BGR
aaaaaaaaa
333231
232221
131211
⋅=BGR
Compute color correction matrix from spectral response, or Macbeth chart image
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 90
Color noise
luminance noise chrominance noise
( ) YB2
YR2
Y2 088.0279.0D −− ++= σσσσ
B0721.0G7154.0R2125.0Y ++= luminance equation
total noise
ISO SpeedFebruary 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors Page 91
Color versus monochrome sensitivity
Monochrome compared to color:•Higher peak QE•Broader spectral response
ISO speed ~ 10 times greater
top related