altering exposure on an image

When merging photographs (e.g. for a panorama)
it is desirable to use a camera where the exposure
can be locked, so that the multiple images
are all taken with the same exposure.

Let's assume one either can't (or more likely didn't)
do this.

How can I correct for either variable shutter speed,
or variable aperture, or both?

I don't think I can just linearly scale
the data, since in a JPEG file from a digicam;
I think the samples are in a log space, with gamma
applied.

So I believe I would need to "ungamma" and "regamma"
before and after any scaling. Does anyone have
actual details? The exif data for my camera
does not have this value.

I plan to use this:
http://netpbm.sourceforge.net/doc/pnmgamma.html

and move into and out of 16 bit representation for accuracy:
http://netpbm.sourceforge.net/doc/pamdepth.html

whilst using this (easy!) for the scaling itself:
http://netpbm.sourceforge.net/doc/pamfunc.html

I think the scaling for aperture is a simple linear
model where quantity of light is linearly
proportional to time, but would welcome confirmation.

Does anyone know how f-stop relates to quantity
of light e.g. what the light level ratio between f5 and f8
would be?

BugBear

bugbear wrote:
When merging photographs (e.g. for a panorama)
it is desirable to use a camera where the exposure
can be locked, so that the multiple images
are all taken with the same exposure.

Let's assume one either can't (or more likely didn't)
do this.

How can I correct for either variable shutter speed,
or variable aperture, or both?
[]
BugBear

Use program like Autostitch and it does it all for you.....

http://www.cs.ubc.ca/~mbrown/autostitch/autostitch.html

What I tend to do for my pano shots (all less than 180 degrees) is to take
exposure from one fixed point (usually the mid point of the set) and,
having set the exposure, swivel to the next image in the set. So I start
with the middle picture, take it, re-expose (half-press), swivel left and
take, revert to the middle, half-press, swivel right and take, and so on.
Easier to do than describe.

Cheers,
David

bugbear wrote:
When merging photographs (e.g. for a panorama)
it is desirable to use a camera where the exposure
can be locked, so that the multiple images
are all taken with the same exposure.

Let's assume one either can't (or more likely didn't)
do this.

How can I correct for either variable shutter speed,
or variable aperture, or both?

I don't think I can just linearly scale
the data, since in a JPEG file from a digicam;
I think the samples are in a log space, with gamma
applied.

So I believe I would need to "ungamma" and "regamma"
before and after any scaling. Does anyone have
actual details? The exif data for my camera
does not have this value.

I plan to use this:
http://netpbm.sourceforge.net/doc/pnmgamma.html

and move into and out of 16 bit representation for accuracy:
http://netpbm.sourceforge.net/doc/pamdepth.html

whilst using this (easy!) for the scaling itself:
http://netpbm.sourceforge.net/doc/pamfunc.html

I think the scaling for aperture is a simple linear
model where quantity of light is linearly
proportional to time, but would welcome confirmation.

Does anyone know how f-stop relates to quantity
of light e.g. what the light level ratio between f5 and f8
would be?

BugBear

With a canon, when shooting in stitch mode exposure and white balance
are set after the first shot. This seemed to work fine for me the few
times I've used this feature. A f-stop step doubles/halves the exposure.
Dave Cohen

Dave Cohen wrote:
bugbear wrote:

With a canon, when shooting in stitch mode exposure and white balance
are set after the first shot. This seemed to work fine for me the few
times I've used this feature.

Yes; my question relates to photographs taken
*without* such features, or equivalent precuations.

A f-stop step doubles/halves the exposure.

Yes, but what about f-numbers?

It appears

http://www.uscoles.com/fstop.htm

that area of the aperture is the key thing,
and that area has a square (or inverse square)
relationship to f number.

Therefore the ratio of f5 to f8 for light gethering

is 8^2/5^2 = 64/25 = 2.56

The numbers for 22 vs 16 = 484/256 = 1.89 which is close enough
to 2 to make me happy.

Can anyone confirm my analysis and/or conclusion?

BugBear

bugbear wrote:
Dave Cohen wrote:
bugbear wrote:

With a canon, when shooting in stitch mode exposure and white balance
are set after the first shot. This seemed to work fine for me the few
times I've used this feature.

Yes; my question relates to photographs taken
*without* such features, or equivalent precuations.

A f-stop step doubles/halves the exposure.

Yes, but what about f-numbers?

It appears

http://www.uscoles.com/fstop.htm

that area of the aperture is the key thing,
and that area has a square (or inverse square)
relationship to f number.

Therefore the ratio of f5 to f8 for light gethering

is 8^2/5^2 = 64/25 = 2.56

The numbers for 22 vs 16 = 484/256 = 1.89 which is close enough
to 2 to make me happy.

Can anyone confirm my analysis and/or conclusion?

I'm no mathematician so I did a spreadsheet and it works like this:
Assume 50mm focal length, it really doesn't matter, then the diameter of
the aperture is fl/f-stop 50/5 & 50/8
so 10 & 6.3
then to get area of the apertu
pi*r^2 or in excel: =PI()*((B2/2)*(B2/2))
so 79 & 31
now the difference in stops is the two areas divided:
1.3 stops difference.

On Tue, 22 Jan 2008 14:01:08 GMT, "David J Taylor"
wrote in
:

bugbear wrote:
When merging photographs (e.g. for a panorama)
it is desirable to use a camera where the exposure
can be locked, so that the multiple images
are all taken with the same exposure.

Let's assume one either can't (or more likely didn't)
do this.

How can I correct for either variable shutter speed,
or variable aperture, or both?
[]
BugBear

Use program like Autostitch and it does it all for you.....

Likewise hugin, which is both excellent and free.

--
Best regards,
John Navas
Panasonic DMC-FZ8 (and several others)

bugbear wrote:
Dave Cohen wrote:

bugbear wrote:
snip
A f-stop step doubles/halves the exposure.


Yes, but what about f-numbers?

Ideally, each f-stop step either halves or doubles the area of the
apeture opening.
This means that the f-stop ratio should go up or down by the square root
of 2 (1.414).
An absolutely "ideal" f-stop series would look something like (if
accurate to 1 decimal place):
f1:1, f1:1.4, f1:2, f2:2.8, f1:4, f1:5.7, f1:8, f1:11.3, f1:16, f1:22.6
and so on.
The only real difference is that there is traditionally a half-stop
between f1:2.8 and f1:4 at f1:3.2 and the numbers are usually rounded to
the nearest whole number higher in series (eg: f1:11 and f1:22).
In practical terms it makes no difference, as film/sensor sensitivity
only steps in whole stops (eg: ISO 100 to 200 to 400) and the small
percentage error caused by rounding f-stop numbers (which may be in
whole stops in the mechanics of the apeture and only marked around the
apeture ring on the lens-barrel in the traditional way) has little or no
effect.

bugbear writes:
When merging photographs (e.g. for a panorama)
it is desirable to use a camera where the exposure
can be locked, so that the multiple images
are all taken with the same exposure.

Let's assume one either can't (or more likely didn't)
do this.

How can I correct for either variable shutter speed,
or variable aperture, or both?

I don't think I can just linearly scale
the data, since in a JPEG file from a digicam;
I think the samples are in a log space, with gamma
applied.

No, the samples are in gamma-corrected space, *not* log space. To a
first approximation, the values in the file are proportional to light
intensity raised to the 0.45 power.

To make the image (for example) one stop brighter if it was in linear
space, you'd multiply all pixel values by 2. In gamma-corrected space,
you multiply all pixel values by 2^0.45 = 1.37. The effect is the
same.

So I believe I would need to "ungamma" and "regamma"
before and after any scaling. Does anyone have
actual details? The exif data for my camera
does not have this value.

For a quick test, you don't need to undo and redo gamma correction; just
multiply the pixel values.

But in reality, the camera encoding process may have used an exponent
somewhat different from 0.45, and it may have compressed the highlights
by using a lower gamma for that portion of the tonal range. For the
highest quality results, you should undo the effects of any such
processing in the camera. But first you'd have to figure out what the
camera did.

I plan to use this:
http://netpbm.sourceforge.net/doc/pnmgamma.html

and move into and out of 16 bit representation for accuracy:
http://netpbm.sourceforge.net/doc/pamdepth.html

whilst using this (easy!) for the scaling itself:
http://netpbm.sourceforge.net/doc/pamfunc.html

If you don't have a model for how the camera encoding departed from the
power-law encoding described above, staying in gamma-corrected space
works just as well.

I think the scaling for aperture is a simple linear
model where quantity of light is linearly
proportional to time, but would welcome confirmation.

Does anyone know how f-stop relates to quantity
of light e.g. what the light level ratio between f5 and f8
would be?

F-numbers are inversely proportional to the diameter of the lens
entrance pupil. Light throughput is proportional to the *square* of
entrance pupil diameter. So to compare light flux for two different
apertures, take one over the square of the ratio. For example, going
from f/5 to f/8 reduces the light by a factor of 0.39.

Dave