ISO Speed Rick Baer February 10, 2005
Outline Introduction Understanding ISO speed ISO speed characterization ISO speed and image processing Summary
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 2
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 Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 3
Understanding ISO speed Exposure controls & metering ISO speed of film ISO speed of a solid-state image sensor
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 4
Basic exposure controls Exposure ~ number of incident photons shutter (period)
sensor
aperture (f#)
ISO Speed February 10, 2005
n.d. filter (density)
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 5
Why control exposure? Dynamic range Shutter speed control Aperture control – depth of field
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 6
Underexposure
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 7
Correct exposure
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 8
Overexposure
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 9
Slow shutter
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 10
Fast shutter
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 11
Wide aperture
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 12
Narrow aperture
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 13
Exposure index
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 14
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 -or2. Select a shutter speed and read the aperture sensor/film exposure index
aperture value
( f # ) 2 EI La = t 15.4 shutter speed ISO Speed February 10, 2005
average scene luminance
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 15
What is the exposure index?
8 EI = Hg
≈
10 Ha
Where: = geometric mean focal plane exposure
used in electronic imaging systems, ISO 12232
= arithmetic mean focal plane exposure Ef (focal plane illuminance) [lux] = ⋅ Te lens ISO Speed February 10, 2005
sensor
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
[lux-seconds] Page 16
Es
Relationship between scene illuminance and focal plane illuminance focal plane illuminance (lux)
scene illuminance (lux)
Ef =
4 ( f # ) (1 + m )
Typical scene illuminance levels: Direct sunlight 100,000 lux Indirect sunlight 10,000 Overcast 1,000 Office 300 Pub 10 Full moon 0.1 ISO Speed February 10, 2005
Es R 2
2
Ef
scene reflectivity (lux) image magnification
Ef ≈ .006 Es For R = 18%, f# = 2.8, m = 0 (1.8 focal-plane lux with office lighting)
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 17
Photopic units
photopic units (lux) describe light intensity as perceived by a human observer
yλ
∞
Φ v = 680 ∫ yλ Φ e (λ )dλ 0
luminous flux (lumens) ISO Speed February 10, 2005
photopic response function
spectral flux (Watts)
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 18
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 exposure Æ 1500 photons/pixel
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 19
ISO speed and exposure index
•The correct exposure for a particular scene determines the exposure index. exposure compensation
•The ISO speed is equal to the exposure index for a statistically average scene.
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 20
High-key scene
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 21
Low-key scene
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 22
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 Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 23
ISO = 200
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 24
ISO = 3200
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 25
ISO speed comparisons are meaningless unless image quality is considered!
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 26
ISO speed of film Speed range of commercial film Speed versus quality tradeoff ISO measurement procedure
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 27
Color negative film
ISO 100
ISO 1600 B&W film
ISO 400 ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 28
Speed versus quality
Film response is nonlinear Film is a threshold detector Grain noise is more important than shot noise Grain size increases with film speed
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 29
Scanned film
(ISO 100) ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
(ISO 800) Page 30
Measurement apparatus calibrated lux meter
variable aperture
shutter speed control camera body (no lens)
Calibrated light source
integrating sphere IR filter
densitometry ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
processing Page 31
Measurement procedure 1. Use aperture and shutter to vary focal plane exposure from frame to frame. 2. Process film for gamma = 0.615 3. Measure film density versus focal plane exposure 4. Plot density versus log exposure 5. Determine exposure intercept at (fog + 0.1) density 6. ISO speed = 0.8 / En
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 32
(film density plot)
ISO = 0.8/En = 0.8/10-2.2 = 125 ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 33
The ISO characterization methods used for film are not applicable to solid-state image sensors, with linear responses and different noise mechanisms.
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 34
ISO speed of a solid-state image sensor Comparison of film and solid-state sensors ISO 12232 methodology Monochrome image sensor model Color image sensor model
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 35
Comparison of film and solid-state image sensors
camera
film
film developer
may incorporate camera functions (e.g. metering, shutter)
camera
image sensor
may be integrated!
image processor
Image sensors and digital processing compare with film and developing ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 36
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 Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 37
vary illumination, shutter speed to control focal plane exposure H
V measure SNR of output signal
sensor test camera
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 38
SNR 40 30 20 10
H10
H40 Hsat
S = 10/H ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
H S = 78/H Page 39
ISO speed model for monochrome image sensors
•Determine H required to achieve specified SNR •ISO speed = 10/ H
photopic response
x
H signal electrons
x QE
noise model
÷
SNR
noise electrons
read noise, dark current ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 40
Photopic quantum efficiency (photopic electrons per lux)
focal plane lux
signal electrons photopic response function
Planckian radiator
Φ e (λ ) = ∞
IR filter
pixel array
−1 α hc / kTλ ( ) e − 1 λ5 ∞
H = 680∫ yλ Φ e (λ )dλ
Ne = ∫
0
0
λ IR (λ ) QE(λ ) Φ e (λ ) hc
dλ
ηp = N e / H ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 41
Typical photopic QE for a monochrome image sensor (Panasonic MN3776; peak QE = 50%) Photopic QE 4.5 4
kiloelectrons/(lux-µ m2)
3.5 3 2.5 2 1.5 1 0.5 0 2500
3000
3500
4000
4500
5000
5500
6000
6500
7000
color temperature [K] ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 42
Noise model
integration time [sec] dark current [e-/sec]
x
sum of powers of uncorrelated noise sources
X2 signal electrons [e-]
= shot noise power
x PRNU ISO Speed February 10, 2005
X2
Σ X2
r.m.s. noise electrons [e-]
read noise [e-]
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 43
(*** neglecting PRNU ***)
Derivation of ISO speed equation Photopic
QE
pixel area
ηp A H S = N x η p A H + Nr 2
focal plane exposure read noise
shot noise
20 η p A 4 Nr 1+ 1+ Sx = 2 2 (S / N) x (S / N ) x 2
−1
•ISO noise speed increases linearly with QE and pixel area •“Acceptable” noise speed (SNR=10) depends on electronic noise ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 44
SNR curves SNR vs. charge 3
Achievable SNR limited by PRNU
2.5
log10( SNR )
2
shot noise only 1.5
NR = 10, PRNU = 0.01 Nr = 10, PRNU = 0
1
0.5
0 0
1
2
3
4
5
6
log10( electrons)
Read noise only affects “acceptable” ISO speed ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 45
ISO speed curves for a typical sensor
(*** neglecting PRNU ***)
ηP = 3.7 ke, Nr =10 e, Nsat = 20 ke Monochrome ISO speed 5000
ISO speed
4000
3000 acceptable excellent saturation 2000
1000
0 2
3
4
5
6
7
8
9
10
pixel pitch [um] ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 46
Effect of IR and UV (wavelengths outside photopic response range) Black body emission 100% 90%
Relative photon flux
80% 70% 60%
far IR near IR visible UV
50% 40% 30% 20% 10% 0% 2000
3000
4000
5000
6000
7000
8000
9000
10000
Temperature [K]
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 47
ISO speed ratio for BB source, w & w/o IR filter Photopic response with IRF / photopic response w/o IRF 3
2.5
Ratio
2
1.5
1
0.5
0 2500
3000
3500
4000
4500
5000
5500
6000
6500
7000
Color temperature [K]
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 48
Effect of dark current In the low-SNR, high noise limit:
Sa ≈ ISO “acceptable” speed
10 η p A Nd Te
dark current [e/sec]
exposure period
ISO speed varies inversely with exposure period: longer integration times don’t help! ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 49
The high ISO speeds of solid-state image sensors (compared to film) are a consequence of the higher QE of silicon.
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 50
ISO speed characterization Camera characterization: ISO 12232 Sensor characterization: QE and noise
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 51
ISO 12232 procedures for determining speed Direct focal plane exposure method Indirect scene luminance method
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 52
Measurement apparatus (direct focal plane exposure) variable aperture
calibrated lux meter shutter speed control camera body (no lens) baffle
Calibrated light source (fixed color temperature)
integrating sphere IR (& UV) filter
raw digital output
Measure SNR versus focal plane exposure ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 53
Illumination control
The angular spread of illumination must match the lens f#.
f# = d/z z
without baffle
Sony ICX252 angular response 120%
F3.5
f2.8 f2.0
f1.4
f0.6
100%
80%
Response [rel.]
d
horz vert
60%
40%
20%
0% 0.00
5.00
10.00
15.00
20.00
25.00
30.00
Incidence Angle [deg.]
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 54
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 Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 55
SNR curve (monochrome CCD) SNR vs. focal plane exposure 60
50
SNR
40
30
20
10
0 0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Focal plane exposure (lux-sec)
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 56
ISO speed from SNR curve SNR vs. focal plane exposure 60
50
SNR
40
30
20
10
0 0
0.01
S10 = 10/0.0044 = 2300 ISO Speed February 10, 2005
0.02
0.03
0.04
0.05
0.06
0.07
0.08
Focal plane exposure (lux-sec)
S40 = 10/0.046 = 218
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 57
Measurement apparatus (indirect scene luminance: reflection) light source color meter luminance meter
scene luminance
shutter speed control
lux meter
camera
(alternative to luminance meter)
lens diffuse reflection target
baffle
(used to adjust luminance)
variac
raw digital output
(used to adjust source color temperature)
Measure SNR versus (scene luminance ⋅ ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
exposure time) Page 58
Measurement apparatus (indirect scene luminance: transmission) variac
color meter
luminance meter
scene luminance
shutter speed control camera lens
light source baffle
diffuser
transmission target raw digital output
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 59
Equation relating focal plane exposure to target luminance luminance
Integration time
65 La t Ha = 100 ( f # ) 2 (1 + m ) 2
Calculate speed as 10/Ha @ SNR = x (as before)
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 60
Need for OECF characterization: perform analysis on OECF target
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 61
SNR curve from OECF target OECF target method 50.0 45.0 40.0 35.0
SNR
30.0 25.0 20.0 15.0 10.0 5.0 0.0 0
2
4
6
8
10
12
patch number
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 62
Sensor characterization: QE and noise QE measurement Noise characterization
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 63
ISO speed calculation from QE and noise (monochrome case) spectral measurements QE(λ) IR(λ) noise parameters noise measurements
ISO Speed February 10, 2005
photopic QE calculation
ηp
noise model
Nr
ISO speed equation
ISO speed range
operating parameters (gain, temp., etc.) 2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 64
ISO speed calculation from QE and noise (color case) color correction matrix calculation
spectral measurements QE(λ) IR(λ) noise parameters noise measurements
ISO Speed February 10, 2005
photopic QE calculation
ηp
noise model
Nr
ISO speed model
ISO speed range
operating parameters (gain, temp., etc.) 2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 65
reference output
QE measurement variable aperture
monochrometer
calibration photodiode
reference photodiode
shutter speed control camera body (no lens) baffle light source low-pass filter
integrating sphere raw digital output Measure QE versus λ
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 66
CC matrix determination
From QE curve From Macbeth chart
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 67
Noise measurement
(see ISO 15739 standard) ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 68
Example: CCD for a typical consumer digital camera Nr = 8 e
50%
40%
Nsat = 10 ke
QE [%]
30%
20%
PRNU = 1%
10%
0% 360
410
460
510
560
610
660
710
760
Wavelength [nm]
Sony ICX202 For 3 um pixel pitch: Acceptable quality speed: SA = 285 Speed range: ~ 50 - 400 ISO Speed February 10, 2005
Excellent quality speed: SE = 23 Saturation speed: Ssat = 72
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 69
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 Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 70
Five (coupled) dimensions of image quality The five “R’s” of image quality Resolution (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 Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 71
Coupling between image quality dimensions
(high resolution, high-noise)
ISO Speed February 10, 2005
(low resolution, low-noise)
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 72
Coupling between ISO speed and other IQ dimensions aRtifacts Reproduction
Resolution
DNR SNR (ISO speed)
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 73
Coupling between ISO speed and other IQ dimensions aRtifacts Reproduction
Resolution
DNR SNR (ISO speed)
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 74
Demosaic Bayer mosaic pattern
σR
σ’R
bilinear reconstruction
σ G′ = 9 / 16 σ G
σ R′ = 5 / 8 σ R
σ B′ = 5 / 8 σ B
Bilinear reconstruction reduces noise (at the expense of resolution) ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 75
Vignetting correction
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 76
Sharpening ~ 30% noise increase
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 77
Tone mapping
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 78
Color space conversion RGB 888 Æ YUV 422, YUV 411 …
Chrominance down-sampling reduces color noise
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 79
Compression
original image
ISO Speed February 10, 2005
JPEG compressed
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 80
Summary ISO 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 February 10, 2005
(ISO speed) × resolution {also true for DNR} ∝ QE × (sensor area) Size matters! 2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 81
Acknowledgements •Albert Theuwissen and the organizers of this forum •Jack Holm
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 82
References 1.
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 Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 83
Appendices •ISO speed model for color image sensors
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 84
ISO speed model for color image sensors
Assumptions: 1. Independent RGB color samples at every pixel location 2. No image processing other than white balance and color correction (e.g. no tone correction, sharpening, compression)
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 85
Extension of monochrome model to color
photopic response x x
H R,G,B signal electrons
QE vector
white balance
noise model
read noise, dark current
ISO Speed February 10, 2005
color correction
R,G,B noise electrons
white balance
RGB to Y
÷
SNR
color correction RGB to Y-cb-cr
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Y-cb-cr to perceived noise
Page 86
Spectral response => photopic QE vector IR filter
5500K black body
SONY ICX224AQ
1.0
100%
40%
0.8
80%
R elative pow er
T r a n s m is s io n
30%
QE [%]
0.6
0.4
10%
0.2
0.0 360
20%
560
660
760
860
360
460
560
660
760
0% 360
860
∞
Ne = ∫ 0
460
560
660
760
860
Wavelength [nm]
Wavelength [nm]
Wavelength [nm]
λ IR (λ ) QE(λ ) Φ e (λ ) hc
ηp = N e / H ISO Speed February 10, 2005
40%
20%
0%
460
60%
dλ
[electrons/lux-second]
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 87
White balance
Calculate coefficients from photopic QE vector
R G/R 0 0 R G = 0 1 0 ⋅G B 0 0 G/B B Apply to noise vector
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 88
Color correction and spectral response SONY ICX224AQ 40%
QE [%]
30%
20%
10%
0% 360
460
560
660
760
860
Wavelength [nm]
Compute color correction matrix from spectral response, or Macbeth chart image
a12
a13 R
G = a 21 a 22
a 23 ⋅ G
a 31
a 33 B
R B ISO Speed February 10, 2005
a11
a 32
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 89
Color noise
luminance noise
Y = 0.2125 R + 0.7154 G + 0.0721 B
chrominance noise luminance equation
σ (D ) = σ 2 Y + 0.279 σ 2 R − Y + 0.088 σ 2 B− Y ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
total noise Page 90
Color versus monochrome sensitivity
Monochrome compared to color: •Higher peak QE •Broader spectral response Æ ISO speed ~ 10 times greater
ISO Speed February 10, 2005
2005 ISSCC Circuit Design Forum: Characterization of Solid-State Image Sensors
Page 91