irees
Member
You're missing the obvious part: cos^4 light falloff correction is NOT dependent on aperture, only the angle of acceptance. Thus, no automatic correction based upon aperture is necessary or even possible.
irees
Member
Personally, I don't give a crap about having the aperture display in the EXIF. My dSLR doesn't record aperture or focal length when I use manual lenses, or even use the built in meter.
jlw
Rangefinder camera pedant
irees said:
Cosine^4 falloff isn't aperture-dependent, but other kinds of optically-induced vignetting are.
Nobody but a few insiders has actually shot with an M8 yet, but we can use as a stand-in the only other digital RF camera that takes M-mount lenses: the Epson R-D 1. What I can tell you by personal experience with an R-D 1 is that most lenses that vignette do so more at wider apertures, so if you're dialing in software-based correction, you'll need different amounts depending on the aperture in use.
Take a look at the two images attached, which I shot just now for the purpose of this discussion. I shot both of them using an R-D 1 and a 50mm f/1.5 Nokton lens (which should be long enough to be reasonably immune from cosine^4 error, wouldn't you think?) In both cases the lens was focused at infinity and aimed at the closest thing I could find around the apartment to an evenly-illuminated white wall. I used the shutter speeds recommended by the camera's TTL meter, then opened and converted the raw files in Photoshop CS2 with all the auto-compensation settings turned off. So, the only difference between the two frames is the taking aperture: f/1.5 for one and f/11 for the other.
The first attachment is the f/1.5 frame and the second is the f/11 frame. Q.E.D., wouldn't you say?
For those who prefer not to believe their eyes, the third attachment shows the histograms of the two images: f/1.5 image on the left, f/11 image on the right. A histogram is a graph that shows the image's brightness range, from dark to light, across the horizontal axis. The vertical axis shows the number of pixels that have each brightness value.
Since the overall tone of both images is medium gray, both histograms consist of a peak in the middle of the horizontal axis. But you can see that in the left histogram, the peak has a wider base -- meaning that there's more brightness variation in the f/1.5 image than in the f/11 image.
Below each histogram, you'll see some statistics about each image including its standard deviation -- a statistical measure of how much the different values differ from the average value. The f/1.5 image's standard deviation is more than 3x that of the f/11 image, again showing that it has much more brightness variation across the frame.
Moral of this somewhat long-winded story: If you're trying to correct for optically-induced vignetting, it helps to know the working aperture of the lens, since the amount of vignetting does vary from aperture to aperture.
Attachments
Bob Ross
Well-known
Those are interesting test images. Another way to do it is with an Expo Disk (or I supposed a Pringle lid).jlw said:
What we may also be seeing is spherical aberration, which should be gone by f/2.8 according to some tests of that lens. There is another aberation where light travels different distances through glass, as the wide aperture angles increase. This has more impact than cosine^4, but is aperture sensative.
If we get similar results with the M8 with a 50mm, we can be happy campers. If we have a problem lens/aperture, we can use an image like yours as a filter in post processing. It would probably do a better job that the uniform correction tools in the software.
Bob
rvaubel
Well-known
jlw said:
I am relieved to find out I am right about something. Your post are generally so well researched that I was afraid I was missing something. I also hope that the blue dot is an external meter as the aperture data would be valuable information to have in the EXIF data. Also it would be nice to see the information in the viewfinder window, especially when using lenses in which the aperture ring rotates the "wrong" way.
Of course white balance info could be gathered for those who are using JPEGs instead of RAW.
willemvelthoven
RD1 user
rvaubel said:
Yes! and even for those who use raw, the white balance measurement could be of value at later time I would expect...
AndyPiper
Established
jlw: right! There are (at least) 3 sources/types of vignetting or corner darkening.
1. Mechanical vignetting - the lens barrel crops the full aperture of a fast lens into a "cat's eye" shape when seen off-center (as in your example). Stopping down crops the "cat's eye" back into a circle as seen from all parts of the film or sensor and evens out the illumination. Aperture dependent - could be fixed by the magic blue dot - if that's what the blue dot is for.
2. Cos-whatever light fall-off. The corners of the image area are further from the exit pupil of the lens than is the center of the image area. Pythagoras, hypotenuse of a right triangle and all that. Light intensity drops off as the square of the distance, so the 'distant' corners get less. Not aperture dependent.
3. Digital vignetting - due to the fact that photo-sites (or pixels) are depressions or pits in the surface of the silicon, and the walls of the pits shade the light-sensitive bottom (just as tall buildings shade a street from low winter sunlight, but not a summer noon sun). Sometimes helped, sometimes exacerbated by microlenses. Depends soley on the angle of the light - which varies slightly with focus, but not at all with aperture.
So 2 of the 3 sources can be corrected simply by knowing the optical formula used - but the third would need to account for the aperture as well.
When Leica tells me what the blue dot is for - then I'll know.
1. Mechanical vignetting - the lens barrel crops the full aperture of a fast lens into a "cat's eye" shape when seen off-center (as in your example). Stopping down crops the "cat's eye" back into a circle as seen from all parts of the film or sensor and evens out the illumination. Aperture dependent - could be fixed by the magic blue dot - if that's what the blue dot is for.
2. Cos-whatever light fall-off. The corners of the image area are further from the exit pupil of the lens than is the center of the image area. Pythagoras, hypotenuse of a right triangle and all that. Light intensity drops off as the square of the distance, so the 'distant' corners get less. Not aperture dependent.
3. Digital vignetting - due to the fact that photo-sites (or pixels) are depressions or pits in the surface of the silicon, and the walls of the pits shade the light-sensitive bottom (just as tall buildings shade a street from low winter sunlight, but not a summer noon sun). Sometimes helped, sometimes exacerbated by microlenses. Depends soley on the angle of the light - which varies slightly with focus, but not at all with aperture.
So 2 of the 3 sources can be corrected simply by knowing the optical formula used - but the third would need to account for the aperture as well.
When Leica tells me what the blue dot is for - then I'll know.
irees
Member
I don't want the camera body to correct for non-cos^4, optical vignetting. This is the silliest idea I've ever heard. I don't know of a single consumer or professional digital body that does it.
irees
Member
#2, inverse squared light falloff is due to the photon flux through a fixed area on the surface of a sphere decreasing with the radius of the sphere... not pythagoras and all that 
Last edited:
rvaubel
Well-known
irees said:
The only lenses where this has a signifcent impact are the very lenses that are traditional for rangefinders i.e. non-retrofocus wide angles. Any lens that has the rear element very close to the image plane is going to have cos^4 vignetting. Most of the newer wide angle rangefinder designs are moving towords the retrofocus designs found in the SLR world. But these lenses are bigger than the older symmetrical designs. And its more difficult to control distortion.
All in all it would be benificial to be able to control vignetting in software. But I would much preferr to do any correction in PS rather than have the camera do it for me. Particullarly since there are tradeoffs with vignette correction. LIke more noise in the corners.
Rex
irees
Member
Yes, I should have been more elaborate in my first post.
Let's classify the light falloff.
Falloff due to CCD angle-of-acceptance issues will not vary due to aperture, because the angle of acceptance doesn't change with aperture. This effect doesn't really exist in film (maybe theoretically due to the film not being infinitely thin and light passing through some film gelatin and chemicals, more at higher angles, but little in practice)
Cos^4 light falloff is different theoretically but has the same basic pattern and dependence on angle as CCD falloff. So, it also won't change with aperture. It will be exactly the same as it would be on film.
Optical vignetting due to the construction of the lens will vary due to aperture, but will behave exactly the same as it would be on film.
(and other minor sources, but no diff. film vs. ccd)
Of these, the only one that is different with CCD vs. Film is CCD angle of acceptance. This is what the 6 bit codes are for. The aperture doesn't need to be known.
Let's classify the light falloff.
Falloff due to CCD angle-of-acceptance issues will not vary due to aperture, because the angle of acceptance doesn't change with aperture. This effect doesn't really exist in film (maybe theoretically due to the film not being infinitely thin and light passing through some film gelatin and chemicals, more at higher angles, but little in practice)
Cos^4 light falloff is different theoretically but has the same basic pattern and dependence on angle as CCD falloff. So, it also won't change with aperture. It will be exactly the same as it would be on film.
Optical vignetting due to the construction of the lens will vary due to aperture, but will behave exactly the same as it would be on film.
(and other minor sources, but no diff. film vs. ccd)
Of these, the only one that is different with CCD vs. Film is CCD angle of acceptance. This is what the 6 bit codes are for. The aperture doesn't need to be known.
Last edited:
ghost
Well-known
how much does reducing vignetting through software increase noise in the corners?
irees
Member
ghost, at least as much as raising the ISO an amount equal to the light falloff. And likely more, because some sensors don't give a truly linear response, and put most of the 12 significant bits in the highlight region instead of the shadow areas.
ghost
Well-known
ah. guess we'll have to be careful about not overdoing it.
rvaubel
Well-known
ghost said:
Practically speaking, it cam be quite a bit. that's one of the reasons I don't want the camera doing it for me. It's kind of a balancing act, has to be done in a case by case basis. However, it is useful if you don't overdo it.
Morever, vignetting control can make a lens that has serious vigentting issues (like my VC 15mm) can be much more useful. My experience has been that serious vignetting problems become managable with software management.
willemvelthoven
RD1 user
I guess there always be manual
I guess there always be manual
Several people have expressed their disliking of any in-camera corrections going on.
I guess that Leica will never introduce obligatory automatic sorrections.
Also I'd expect RAW to remain RAW; all this correction stuff is for JPEG shooting...
(and maybe for presetting RAW conversion software)
I guess there always be manual
Several people have expressed their disliking of any in-camera corrections going on.
I guess that Leica will never introduce obligatory automatic sorrections.
Also I'd expect RAW to remain RAW; all this correction stuff is for JPEG shooting...
(and maybe for presetting RAW conversion software)
LCT
ex-newbie
This blue dot makes me think of laser rangefinders like that of the Leica Geovid binoculars.
See what i mean?
See what i mean?
semrich
Well-known
If you check image 3 0f 7 (Mise à jour 07/09/13h00
Pièce révélée: 09) on the M8 puzzle you may notice that the "blue dot" is red or at least has a red border around it.
http://www.lightmediation.com/blog/index.php?2006/09/06/25-leica-m8-digital-numerique-exclusif
Richard
Pièce révélée: 09) on the M8 puzzle you may notice that the "blue dot" is red or at least has a red border around it.
http://www.lightmediation.com/blog/index.php?2006/09/06/25-leica-m8-digital-numerique-exclusif
Richard
Leicaiste
Member
Easy !
Don't use 6 bits lenses if you don't want corrections.
Don't use 6 bits lenses if you don't want corrections.
Jon Graham
Established
This whole blue dot thread makes me want to lie down in a darkroom as Mark Norton suggested.
At the risk of sounding very stupid why cannot the in camera software roughly calculate the aperature from the following:
-shutter speed
-Iso
-Histogram
I really don't know WTF the histogram can do to solve the equation - but we are dealing with a digital animal - not a film camera. There is in camera processing.
At the risk of sounding very stupid why cannot the in camera software roughly calculate the aperature from the following:
-shutter speed
-Iso
-Histogram
I really don't know WTF the histogram can do to solve the equation - but we are dealing with a digital animal - not a film camera. There is in camera processing.