Mail
NEW

Number 30

Disclaimer: All the below text are my personal thoughts, so they don't have to reflect the thoughts of the people that made the formula/guidelines.

I have the stereophographic guidelines also as functions (Add-in) for Excel available, so one can build one's own application around this Excel Add-in. If you want more information, let me know!

Stereo glossary: some definitions

Purple definition is not used on my web site, but given because other people sometimes use these definitions.

Term	Definition/description
3D image	the image is the virtual 3D scene that is seen when looking at a stereo pair, this is experienced with the brain.Several types are defined: geometric, picture size, immersion Sometimes called: cyclopean image
3D scene	The actual scene in real live (make up of physical objec ts)
a'_c	lens to film/chip distance [mm] (1/a'_c=1/F_c-1/focusdistance)
nearestobject_act	distance nearest object in actual scene [mm]
nearestobject_exp	distance nearest object in experienced image [mm]
farthestobject_act	distance farthest object in actual scene [mm]
farthestobject_exp	distance farthest object in expereinced image [mm]
b0	stereo base [mm] of camera
CoC	Angular Circle of Confusion [°], sometimes also in [mm]
con/divergence	see vergence
deviation	The 2D displacement [mm or pixel] of homologous points in the stereo pair, which is a result of a certain depth in the 3D scene. This term is equivalent with OxD. The deviation is the difference between the 'homologous separation of the point under investigation' and the 'homologous separation of the farthest points'. (equivalent to Ferwerda [1990], page 236-237) Deviation can be in the horizontal of vertical direction (or a combination of the two, like a vector in a (2D) plane). (some people call this parallax [mm or pixel], which seems to be true in astronomy for instance and in technical paper it is also called disparity).
D_H	Hyperfocal distance, this is closest distance that is perceived as sharp when the camera lens is configured for infinity [mm]. (Hawkins [1980], page 71)
eye separation	The distance between the two optical axis of the eyes [mm] (some people call this Inter Pupil Distance: IPD, Inter Ocular: IO)
filmwidth	width of uncropped film/chip [mm or pixel]
F_c	camera's focal length [mm]
FoV	Field of View [°] is the horizontal angular part visible in the actual (_act) scene or experienced (_exp) image.
F_v	stereo viewer's focal length [mm]
fixation distance	equivalent to vergence distance, but fixation is used for distance/angles related to the human eye.
focusdistance	The focused distance [m] Accomodation is also used for the focused distances of the eye.
HAOxD	Highest Allowable On x Deviation [mm or pixel] (threshold deviation related to highest comfortable experienced parallax angle [~4°]). x can be film/chip (F), screen/display/picture (P), viewfinder (V), retina (R), etc
LAOxD	Lowest Allowable On x Deviation [mm or pixel] (threshold deviation related to parallax angle which still gives enough experienced stereoscopic depth [~1°]) x can be film/chip (F), screen/display/picture (P), viewfinder (V), retina (R), etc
M	Magnification (a'_c/focusdistance) [m]
MAOxD	Maximum Allowable On x Deviation [mm or pixel] (I think this is the same as my OOxD...) x can be film/chip (F), screen/display/picture (P), viewfinder (V), retina (R), etc
MOxD	Actual Maximum On x Deviation [mm or pixel] (due to nearestobject_act and farthestobject_act), same as NetOxD x can be film/chip (F), screen/display/picture (P), viewfinder (V), retina (R), etc
NetOxD	Netto On x Deviation [mm or pixel]. This is the same as my MOxD. x can be film/chip (F), screen/display/picture (P), viewfinder (V), retina (R), etc When people talk about NetOFD they determine it by: NetOFD=deviation_winnear - deviation_winfar with: deviation_winnear: the deviation of to nearest object relative to the deviation of the stereo window [mm or pixel] deviation_winfar: the deviation of to farthest object relative to the deviation of the stereo window [mm or pixel] A deviation_win in front of stereo window is negative, while behind it it is positive.
nonstereooverlap	non stereo overlap on film/chip/screen [mm or pixel] at the stereo window (is normally cropped away).
OxD	The actual On x Deviation [mm or pixel] (due to a certain homologous distant point and farthestobject_act point). Same as deviation x can be film/chip (F), screen/display/picture (P), viewfinder (V), retina (R), etc
OOxD	Optimum On x Deviation [mm or pixel] (related to an optimum experienced parallax angle [parallax_opt~1.9°]) x can be film/chip (F), screen/display/picture (S), viewfinder (V), retina (R), etc.
parallax	parallax angle due to different standpoints (like due to eye separation or stereo base) for a certain point in a 3D scene [°] (Ferwerda [1990], page 237) Parallax can be an angle in the horizontal of vertical plane (or a combination of the two, thus like a vector in a 3D scene). Some people use the word parallax instead of deviation.
parallax_opt	optimum parallax angle of 1.9 [°] for instance due to a 2 m nearest object, farthest object at infinite and an eye separation of 65.5 [mm]
picture	is a 2D representation of the 3D scene.
Q	a factor between filmwidth and deviation [-]
homologous separation	The distance between homologous points in the two pictures of the stereo pair [mm or pixel].
shifting	Shifting the pictures of the stereo pair in horizontal or vertical direction relative from each other [mm or pixel] (the horizontal shifting changes the experienced stereo depth of the stereo window). (some people call this parallax [mm or pixel], this is confusing, because the shifting has nothing to do with parallax [aka. different angles])
stereo base	The distance between the two object sof the camera [mm] (Some people call it Inter Axial: IA)
stereo depth	The distance between nearest and farthest object [mm] in actual life or experienced image
stereo pair	consists of two pictures taken from different viewpoint (distance between viewpoints is stereo base)
stereo window	the (2D) plane of the 3D image that is experienced at the same depth as the display screen (display, projector screen, paper, frame, etc.). At this plane there is normally a 100% stereo overlap for the objects that are at that particular depth/distance. If the homologous separation of certain points is the same as the frame separation, the stereo window is at the frame plane.
vergence angle	The angle between optical axis of the two lenses/eyes. A vergence angle of zero meas parallel axis, a positive vergence angle is convergence and a negative vergenbce angle is divergence Some people use half it (so defined towards the parallel line)
vergence distance	The vergence distance tells where the optical axis of the two lens systems cross (under vergence angle). If this distance is in front of the lens system we talk of convergences, if behind it is divergence.

Another glossary site is here <a very extensive one> and the one of ISU <this one is not really consitent enough for me, that is why I made the above one>.

Relations between photographic (stereo) parameters

The diagram gives an idea of the relations between photographic (stereo) parameters.
So for instance the stereo base (b0) is determined by the MOFD (Max On Film deviation), the Film to lens distance (a'_c), the actual nearest and furthest object distances (and if one wants to utilize the Davis correction [in case furthest distance is smaller then 2 times the nearest distance], although I think one needs to change experienced nearest and furthest object distances instead).
Furthermore one can see that the stereo base (b0) is used for determining the nonstereo overlap, convergence distance and number of depth steps. By changing e.g. MOFD (through changing experienced nearest object distance), the number of depth steps will also change.

The color codes have the following meaning:

yellow
The reason why we do it; to determine the stereo base.
green
parameters which are input values, these values are determined by the wishes of the stereo photo maker (when changing the composition and/or quality of the stereo picture will change)
while
the parameters calculated from the green values. The functions to calculate them are available as JavaScript and as an Excel Add-in.
blue
check that indeed the actual diaphragm is the same as the target one, otherwise the picture won't have the proper view of depth. For orthostereo, experienced and actual nearest (and furthest) object distance should be the same.
gray
parameters that one also needs to check and see if they are as wished.
purple border
parameters specific for stereo photography

Overview of rules/guidelines and their conditions

The meaning of the condition color coding (and rules/guidelines as a result of the conditions):

red
Watch out, a condition is present so that the rule/guideline only works in certain instances
dark orange
A condition is present that is not that serious. Normally oke, but still watch out
light orange
A condition that is normally not important (but in certain instances can be important)
white
No condition applied (above Bercovitz Case 1)

Rule/guideline

Condition

farthestobject_act
<=2*nearestobject_act

farthestobject_act>
100*nearestobject_act

deviation
=F_c/30

deviation
=a'_c/30

deviation
=1

deviation
=1.2

b0/nearestobject_act=
1/30

25<=
Filmwidth/MOFD
<=35

Filmwidth/MOFD

=30

F_c<<
2*nearestobject_act

F_c
=35

Stereo base

Bercovitz Case 1

Di Marzio1

Di Marzio2

Di Marzio3

Sekitani

Herbig

Reijs

Deviation

Ferwerda

Di Marzio4

Di Marzio5

Non stereo overlap

DrT

Ferwerda

Miscellaneous

Lopp

ISU

PKK

Glickman

Below the formulas and guidelines are given for stereo photography (not for stereo macro stereophotography).

Stereo base

Bercovitz's stereo base formula

Case 1 of Bercovitz
b0=deviation/(farthestobject_act-nearestobject_act)*(farthestobject_act*nearestobject_act/F_c - (farthestobject_act+nearestobject_act)/2)

Di Marzio1's stereo base formula

Case 1 of Bercovitz
take: deviation=F_c/30
b0=D_H/60 (The Di Marzio equation for stereography, Frank Di Marzio, 2005, page 3)

Di Marzio2's stereo base formula

Eqn 1 of Bercovitz:
take: m=farthestobject_act/nearestobject_act and m>=2b0=deviation*nearestobject_act*m/(m-1)/a'_c
take: deviation=a'_c/30
b0=nearestobject_act*m/(m-1)/30 (The Di Marzio equation for stereography, Frank Di Marzio, 2005, page 19)

Di Marzio3's stereo base formula

Eqn 1 of Bercovitz
assume: farthestobject_act<=2*nearestobject_act (shallow objects)
take: farthestobject_act=2*nearestobject_act (David's condition)
take: deviation=a'_c/30
b0=nearestobject_act/15 (The Di Marzio equation for stereography, Frank Di Marzio, 2005, page 14)

Sekitani's stereo base calculator formula

Case 1 of Bercovitz:
take: F_c<<2*nearestobject_act
b0=deviation*nearestobject_act* farthestobject_act/ F_c/( farthestobject_act- nearestobject_act)

Herbig's stereo base guideline

Case 2 of Bercovitz (farthestobject_act>100*nearestobject_act)
take: F_c<<2*nearestobject_act
b0=deviation*nearestobject_act/F_c
take: deviation = 1 mm (which is close to the OOFD=1.2 mm as used with 36*24 mm film)
b0=nearestobject_act/F_c

Reijs' stereo base and deviation formula

Case 1 of Bercovitz:
As much as possible is open, including the deviation (although that one is coupled to the deviation seen on the retina)
The deviation can vary considerable because the human eye can cater for a large parralax range when stereo viewing; between LAOFD and HAOFD.

To visualize this rule for a particular configuration (just as an example):
Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
take: F_c<<2*nearestobject_act
deviation= F_c*b0/nearestobject_act
take: b0/nearestobject_act=tan(parallax)
deviation(=MOFD)= F_c*tan(parallax)
guideline: Q=filmwidth/MOFD and filmwidth is cropped filmwidth without non-stereo overlap
Q=filmwidth/F_c/tan(parallax)
tan(parallax)=filmwidth/F_c/Q
take: filmwidth = 36 mm
Looking at a parallax between 1 and 4 degrees (related to LAOFD and HAOFD) gives a more or less comfortable/allowable stereo viewing. So the greenish areas in the below picture (and even the yellowy area, although in that area less stereo is experienced) provide a space with allowable stereo experience.
The below picture has F_c as 35 mm equivalent focal length (but equivalent F_c can be calculated for other [digital] cameras).

VR's guideline

Deviation

Ferwerda's deviation guideline

Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
take: b0/nearestobject_act=1/30
take: F_c=35 mm
deviation(=OOFD)=1.2 mm (for 5-P: 23 mm chip)

Di Marzio4's base deviation guideline

Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
b0=deviation*nearestobject_act/a'_c
deviation=a'_c*b0/nearestobject_act
take: b0/nearestobject_act=1/30
deviation_base=a'_c/30 (he multiplies deviation_base with a Correction Factor [CF] depending on viewing method)

Di Marzio5's base deviation guideline

Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
take: F_c<<2*nearestobject_act
b0=deviation*nearestobject_act/F_c
deviation=F_c*b0/nearestobject_act
take: b0/nearestobject_act=1/30
deviation_base=F_c/30 (he multiplies deviation_base with a Correction Factor [CF] depending on viewing method)

Non stereo overlap

DrT's non stereo overlap formula

Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
take: F_c<<2*nearestobject_act
nonstereooverlap(=deviation)= F_c*b0/nearestobject_ac(nonoverlap=FB/I)

Ferwerda's non stereo overlap guideline

Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
take: F_c<<2*nearestobject_act
nonstereooverlap(=deviation)= F_c*b0/ nearestobject_act
take: b0/ nearestobject_act=1/30
nonstereooverlap= F_c/30

Miscellaneous

Lopp's b0/nearestobject guideline

Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
take: deviation(=OOFD)=1.2 mm (as used with 5-P, 7-P or 8-P [36*24 mm] film)

F_c	b0/nearestobject_act
35	1/30
50	1/40
74	1/64
135	1/116

ISU guideline

filmwidth/MOFD=30

Remark: one has to watch out using this definition as it mkaes the MOFD depending on the filmwidth, which is in prinicple not true. It is the absolute value of MOFD [mm/pixel] that is important.

PKK guideline

Q=croppedfilmwith/MOFD
35>=Q>=25

Remark: one has to watch out using this definition as it mkaes the MOFD depending on the filmwidth, which is in prinicple not true. It is the absolute value of MOFD [mm or pixel] that is important.

To visualize this rule for a particular configuration (just as an example)
Case 2 of Bercovitz: (farthestobject_act>100*nearestobject_act)
take: F_c<<2*nearestobject_act
deviation= F_c*b0/nearestobject_act
take: b0/nearestobject_act=tan(parallax)
deviation(=MOFD)= F_c*tan(parallax)
guideline: Q=croppedfilmwidth/MOFD and cropped filmwidth is without non-stereo overlap
Q=croppedfilmwidth/F_c/tan(parallax)
tan(parallax)=croppedfilmwidth/F_c/Q
take: croppedfilmwidth = 36 mm

The Q is checked realtime in the viewfinder. Keeping Q between 25 and 35 by moving around the camera (by changing F_c, b0 and/or nearest/farthestobject_act), this guideline can work because it utilizes the flexibility of the parallax that can be experienced by the eye/brain combination (blue rectangular in below picture, which is superimposed on Reijs' guidelines).

70" rule retinal parallax guideline

max. retinal parallax = ~1.2 deg

According to Ferwerda (page 218, [1990]) the reasoning around this idea is wrong.
The reasoning was that when viewing in a stereoscope, the eyes accommodate at infinite (which is correct). Due to the assumed strong relation between accommodation and convergence, they assumed no convergence of the eyes was possible (as accommodation was at infinite) (this assumption is wrong*).
Due to DoF, accommodation at infinite will still give focused subjects at ~3 m. This 3 m was seen as the limit of retinal parallax: which is thus equivalent to ~70".

*as at 3 m distance (even in stereoscope viewing) convergence will happen, even if accommodation of eyes is at infinite.

Piper's projection guideline

max. retinal parallax = ~1.2 deg

Glickman retinal parallax guideline

max. retinal parallax = 1.3 deg

A retinal parallax of 1.3 deg is proposed by Glickman (pers. comm. [2009]):
His reasoning: When looking at the nearest object in a stereo pair, the farthest object will not shift when the eyes are forced to converge on the nearest object (as the stereo pair is static). When viewing an converging into a real 3D scene, there will be a shift.
This conflict can be removed is by limiting the overall viewing parallax angle to approx. 1.3 deg or less... as at approx. this viewing angle, no physical eye convergence takes place:
This lack of convergence is a result of foveola FoV of approx. 1.5 to 2 deg, which allows s leeway before physical convergence will take place.

Reijs' evaluation of above guidelines

In a lot of guidelines/formulas; the factor 1/30 pops up which is in my opinion related to tan(parallax_opt) and parallax_opt = 1.9 degrees (pink line in below picture). The PKK guideline (the blue rectangle) utilizes already more flexibility of the parallax angle.
Looking at a parallax between 1 and 4 degrees (related to LAOFD and HAOFD, as used in Reijs' stereobase.) gives a more or less comfortable/allowable in stereo viewing. So the greenish areas in the below picture (and even the yellowy area, although in that area less stereo is experienced)) provides a space that provides stereo experience in an allowable form.

PKK's guideline

Disclaimer and Copyright

Home

Mail

Major content related changes: May 18, 2005