Re: Primer on photographic exposure, etc.
Re: Primer on photographic exposure, etc.
- Subject: Re: Primer on photographic exposure, etc.
- From: Thomas Lianza <email@hidden>
- Date: Tue, 21 May 2013 11:25:48 +0000
- Thread-topic: Primer on photographic exposure, etc.
Hi to all,
"Another yet connected thing is that RGB about 30 (on 0..255 scale,
gamma 2.2) is too "light" for what cameras can do now."
In a gamma 2.2 level 30 in an 8bit scale represents about 1 percent of
full scale. This represents a little more than a real-life outdoor scene
without speculars. (That is why video and sRGB systems produce pleasing
images) Now it is true that one can easily find scenes that exceed this,
but the fact is that they normally would have an image of the sun or
reflected image of the sun in the scene to exceed this value. At the
other end of the scale, looking into deep forest will extend the range,
but that range is limited by the atmosphere which on the average limits
the range between diffuse highlight and deep shadow to about 7 stops even
without camera flare. The maximum dynamic range that we can get out of a
matte finish target is approximately 60 to one, which falls just within
the 6 stop range.
The discussion about changing exposure is just too broad to make any
realistic conclusions. There seems to be misconception that changing ISO
is simply affecting an amplifier in the chain. The fact is that it is not
that simple. Based upon the required exposure and anticipated shutter
speed, one may choose to increase the base integration time of the sensor,
within the limits of the shutter speed. Under these conditions, the zero
level can become very unpredictable.
When I designed the exposure checking patches in the color checker
passport, I started with the diffuse highlight and moved 1/3 stop down the
scale. At the lower end, I moved up the scale 1/3 stop(except for the last
two steps, because I couldn't get the darkest patch dark enough). A good
experiment for those that really care to understand how their camera is
actually working would be to modify exposure using the ISO scale, fixing
shutter speed, and compensating using f stop. A second test is to fix ISO
and change f number and shutter speed. If you do this test in a
controlled environment, the diffuse highlight should never change in
value, but you may be surprised about what is happening in the darkest two
patches on that target. In theory, they should track completely with the
diffuse white. In practice, that is rarely the case. You don't need a
passport to do this test, any target with a full white and full black can
be used. The only reason to use the passport is that you have the 1/3
stop targets in the shadow.
It is also important to point out that "Raw" data is rarely, if ever,
truly raw. No manufacturer in their right mind would export truly raw
sensor data. If you read the Canon patents, you begin to see that they
take great pains to make crap sensors useable in the camera on a routine
basis. What you see in a raw file is heavily processed data to minimize
small gamma artifacts, hot pixels, and dark noise. There may also be a
subtle flare compensation in the data chain. Modern cameras can detect
the lens and can apply some lens profiling under the hood. The leica m9 is
an example that uses explicit lens coding to modify the raw data.
If you want to use your camera as a colorimeter, you need to fix exposure,
fix the lens, and the lighting. You will need to build a 3d LUT, because
there is no way to fix flare and other non-linearities in the sensor in a
general fashion. If you are a fine art photographer, your target should
reflect (pardon the pun) the media you are photographing. Monochromaters
can only get you so farŠ.I would also add that if you are using the color
checker, the violet patch, should be avoided in profiling activities.
This patch was designed to be very sensitive to different lighting
situations. It was designed to really test film response and simulates a
naturally occurring spectra in flowers that is nearly impossible to
reproduce in film.
Tom Lianza
On 5/21/13 3:46 AM, "Lars Borg" <email@hidden> wrote:
>>BTW, shooting targets filling the frame with the target may be not
>>the best approach.
>
>No, don't do that. (flare, vignetting...)
>Or establish the light distribution by also shooting a gray card
>overlaying the chart, then normalize the light.
>
>
>>Profiles (especially matrix) can be multiplied.
>
>That's probably not accurate for narrow spectra, but may work OK for
>very smooth spectra.
>
>>Actual light is an extremely interesting question, I do not feel
>>like I can do better with actual light than just measuring it with
>>some portable spectrophotometer.
>
>Right, but then you have to redo the spectral math to get the matrix.
>
>>Noise.
>
>Noise can be averaged, by adding more pixels, multiple chart shots.
>
>>In-camera flare and lens flare.
>
>Yes. That's a challenge.
>While you may be able to subtract much flare from the camera data
>during calibration, that doesn't help in actual color management as
>the flare is not removed from images before conversion.
>As an alternative add estimated flare to your reference data,
>matching how the chart looks to the camera, not how it measures.
>
>>Another yet connected thing is that RGB about 30 (on 0..255 scale,
>>gamma 2.2) is too "light" for what cameras can do now.
>
>Don't quite follow you. Yes, the sRGB tone curve has a limited
>dynamic range and shadow resolution.
>For comparison, digital cinema projection uses 12-bit with gamma 2.6.
>Supposedly that's enough to not show any artifacts for a 2000 : 1 DR
>at 48 nits.
>
>Lars
> _______________________________________________
>Do not post admin requests to the list. They will be ignored.
>Colorsync-users mailing list (email@hidden)
>Help/Unsubscribe/Update your Subscription:
>m
>
>This email sent to email@hidden
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Colorsync-users mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden