1. Mesopic vision model and its application
János Schanda
Virtual Environments and Imaging Technologies Laboratory
University of Pannonia

2. Overview Mesopic vision fundamentals
The five photosensitive cells in the human retina
Luminance type and brightness type description
The CIE TC 1-69 mesopic model
Examples of application

3. Luminance levels

4. Mesopic vision Classical interpretation Daylight: photopic – cones
Dark adaptation: scotopic – rods
Twilight vision: mesopic – cones + rods
Present day knowledge
Foveal vision: photopic
Pupil diameter: intrinsically photosensitive Retinal Ganglion Cells (ipRGC)?
Difference between perception and detection

5. Spectral responsivity of light sensitive cells in the human retina
3 types of cones, rods and ipRGCs (Cirk.-Gall) 5

6. Perception and detection
seeing details, perceiving brightness
all 3 cone types & rods
+ ipRGC (?)
slower
Detection:
only L & M cones + rods luminance like signal
fast

7. Colour perception

8. Historic overview The two stumbling blocks of mesopic photometry:
Purkinje shift
rod and cone interaction
transition from cone to rod sp.resp. differs from Purkinje shift
Helmhotz-Kohlrausch effect
difference between luminance and brightness 8

9. Early investigations Fovea: only cones Peripheric vision: rods + cones
Luminance like: rapid, contrast
Brightness + colour: slower mechanism
Peripheric vision: rods + cones
In mesopic the influence of rods increases

10. Early investigations Brightness description:
Kokoschka 3 conew + rods
Sagawa brightness model
Contrast threshold investigations
Non-linear!
Reaction time based models
aV(l)+(1-a)V’(l)

11. Spectral responsivity at different luminance levels
Walters & Wright : Proc. Roy. Soc
At low mesopic levels first a shoulder is found at longer wavelength
At mid-mesopic a transition from rod to cone takes place: broadening of curve
At high mesopic narrowing to V(l) is found

12. Spectral responsivity at different luminance levels
Kinney clearly identifies in 1955 that at high mesopic and photopic levels brightness sensitivity has a multi bump nature

13. CIE 1963 mesopic spectral sensitivities

14. Brightness perception
Observation
Coloured lights brighter that white (or yellow)
Influence of
S cones
Rods, even in daylight
ipRGC, responsible also for the circadian rhythm

15. Quantitative descriptions of mesopic luminance
Equivalent luminance (CIE 1963): „The equivalent luminance of the field of an arbitrary spectral composition is the standard luminance of another field which has the colour temperature of 2042 K and which in particular photometric conditions seems to be equally bright to the first field.”

16. Mesopic 2 function brightness scale
Palmer (1966, 1967, 1968):
Equivalent luminance (L) for large fields
L(S,P)=(MS+P 2)/(M+P)
where:
S: scotopic value
P: photopic (10°) value
M parameter, cd/m2 for 15° field
Other models used direct cone brightness functions (e.g. Ikeda & Shimozono, 1981; Nakano & Ikeda, 1986; Sagawa & Takeichi, 1983; modified Palmer formula)

17. Mesopic 4 function brightness
Helmhotz-Kohlrausch effect: brightness non-additivity, influence of all three cone types
Kokoschka model
Trezona – Clarke model
Problem of transition between luminance type (photopic) and brightness type (scotopic) experimental data: Viénot et al.

18. Kokoschka model where Fx, Fy, Fz functions depend on luminance level,
L10 is photopic 10°luminance
S is scotopic luminance
X10, Y10, Z10 are 10° tristimulus values

19. Mesopic models Lighting Research Center of North America system:
with 0,001 cd/m2 < Lmes < 0,6 cd/m2
MOVE model, based on
Ability to detect target
Speed of detection
Ability to identify details of target
with soft transition to scotopic and photopic at
0,01 cd/m2 < Lmes < 10 cd/m2

20. Kokoschka model functions and derived spectral luminous efficiency functions

21. Brightness/luminance discrepancy & rod/cone interaction + ipRGCs
In photopic regime one assumes:
luminance to be a combination of the L- and M-cone signals
Brightness to be a combination of all three cone signals after a transformation into magno-, parvo- and conio-cellular signals
In mesopic regime rod contribution has to be added.
Recently found ipRGCs have influence on pupillary reflex, influencing light reaching the retina

22. Photopic regime: brightness/luminance
Magnocellular pathway: luminance like
Brightness: all 3 channels
B=(L2+d2+t2)1/2
(Guth model)
Influence of ipRGCs? Different brightness of metameric samples

23. Mesopic: rod contribution
Two pathways for rod-cone interaction
Classical: via rod bipolar (RB) and amacrine (RA) cells to cone bipolars (DCB & HCB)
Direct pathway via gap junctions
From Buck SL: Rod-cone interaction in human vision, The visual neuroscience

24. 2 new CIE reports CIE supplementary system of photometry, CIE 200:2011
Recommended system for mesopic photometry based on visual performance, based on MOVE and X-models

25. Detection Traffic situation Can be approximated by and additive system
Detecting the presence of an obstacle
Rapid action necessary
Can be approximated by and additive system
Abney’s law holds photometry possible
Should have smooth transition to photopic and scotopic at the two ends.

26. Comparing the two systems
Two lamps with S/P ratio: 0.65 and 1.65: difference of mesopic lum to photopic lum. in the two systems

27. CIE TC 1-58 system, 1
Compromise solution between the two experimental systems, main input data:
achromatic contrast
reaction time (see ball in windshield of virtual reality simulation)

28. CIE TC 1-58 system, 2 The system is not for visual performance :
if chromatic channel signals are important: S/P ratio very higy or low
if target has narrow band spectral power distributions
if brightness evaluation is required
Mesopic limits: 0,005 cd/m2 < Lmes < 5 cd/m2
The TC 1-58 system is for adaptation luminance, i.e. background luminance, not for calculating mesopic luminance of target
Foveal vision is photopic!

29. Calculating mesopic luminance, 1
Photopic luminance Scotopic luminance
Mesopic luminance:
where
and
Vmes(l0)=Vmes(555nm)
m =1 if Lmes>5.0 cd/m2
m =0 if Lmes<0.005 cd/m2
M(m) is a normalizing constant: Vmes,max=1

30. Calculating mesopic luminance, 2
m is calculated using iteration
Start with m0=0.5
Calculate Lmes,n from Lmes, n-1:
where

31. Vmes at different m values

32. Calculation from pavement illuminance
Input data:
Photopic luminance: Lp
Luminance coefficient of road surface (q=L/E)
S/P ratio of light source, where
and S(l) is the rel.sp. power distribution (SPD) of the lamp to be used

33. Calculation from pavement illuminance
Calculate Lp=qE
Calculate S/P and with Lp determine Ls
Calculate Lmes,1 from with m0=0.5
And do the iteration, usually 5 to 10 iterations are needed to get final Lmes
If Vmes is required, used

34. Some examples q= 0.0016 and q= 0.032 Typical light source S/P values:
LPS
0,25
HPS
0,75
LED-2700K
1,12
LED-4000K
1,91

38. Numeric evaluation

39. Visual acuity and lamp spectrum
Transmission of eye media changes with age
Test with cool-white and warm-white LEDs
Young observers: < 30 years of age
Old observers: > 65 years of age
Reading Snellen table at 0.1 cd/m2 and 1 cd/m2

40. Visual acuity and lamp spectrum, results
Young observers have less errors at 0,1 cd/m2 under CW-LED
At 1 cd/m2 the difference is not significant

41. Summary
The mesopic photometry model is valid for background adaptation luminance
It refers to reaction time type of tasks, not brightness
For foveal vision V(l) based metric (photopic photometry) is valid!
It is an experimental model for trial, has to be validated with real street lighting tests and accident simulations
In preparing new recommendations spectral vision differences between young and old observers should be considered

42. Peter Kaisers vision of the future photometer in 1981
Modern image taking photometers are almost there
+ info:
background
ipRGCs