Re: RPP raw photo processor 64
Andrew - I'm just catching up on this thread and have found it interesting. Your response to Ben about white balance on the beach at sunset causes me to ask if there's some "native reference" or starting point for white balance. What happens when a white balance is made? I know what it attempts to do but I don't understand how it goes about it. Is there a null white balance or table of some sort or is there a common one assumed such as daylight and the new balance adjusts that? You're right that the cast would get screwed up if white balance is applied to the beach scene. Prints made from color neg film were all over the place with this kind of scene but some were actually pretty darn good. Was it strictly a difference in the creativity of the photographer and print maker or was there a technical approach that could be repeated with accuracy? This business about the illuminant must be central to colorimetric issues and it would be interesting to know more about it. It seems to me that when filters/sensors are designed there must be assumptions made that characterize the incoming light and this can't be gotten around when trying for colorimetric matches. I believe it would take a spectral approach for colorimetric matching but it wouldn't be possible to make a print that verified it. This would only be verifiable by the numbers - not visually. Well, that's what I believe anyhow. Ben - I read your statement differently I guess, and so my answer would be: the camera imposes an inherent error. The colorimetric problem starts there, in the camera. I don't know if a sensor or combination of filters can be made that would make a colorimetric match possible. On the other hand, they sometimes do a pretty good job of matching paint at the paint store. Henry Message: 7 Date: Sun, 02 Jun 2013 10:29:24 -0600 From: Andrew Rodney <andrew@digitaldog.net> To: "colorsync-users@lists.apple.com List" <colorsync-users@lists.apple.com> Subject: Re: RPP raw photo processor 64 Message-ID: <55224C4F-E1E0-4F46-A9E0-0507CE268C27@digitaldog.net> Content-Type: text/plain; charset=us-ascii <Snip> Ben:
Improper white balance and exposure are colorimetric failures
Andrew: Well it certainly isn't a recommend photographic workflow to improperly white balance and improperly expose! If I white balance a scene of a model on the beach at sunset, is that colorimetric failures? It would certainly produce an image that doesn't look anything like the original scene.
On Jun 5, 2013, at 12:53 PM, Henry Davis <davishr@bellsouth.net> wrote:
Your response to Ben about white balance on the beach at sunset causes me to ask if there's some "native reference" or starting point for white balance.
Cameras are (essentially) linear devices. This means that you can consider each channel independently and determine what linear multiplier is necessary to ensure that a piece of Spectralon (e.g.) gets rendered, in that channel, as 100%. If you do that for all three channels independently, when you re-combine the channels into an RGB image, both white balance and exposure have been properly normalized. More here, including pictures and how to simulate the process yourself in Photoshop: https://trumpetpower.com/photos/Exposure#Normalizing_exposure You'll notice that, although that's whats going on behind the scenes, only the ``alternative'' raw development applications give you the option to specify those channel multipliers for yourself. Everything else, especially ACR, instead invisibly translates blackbody radiation sliders into those channel multipliers using some secret sauce recipe.
[T]he camera imposes an inherent error. The colorimetric problem starts there, in the camera. I don't know if a sensor or combination of filters can be made that would make a colorimetric match possible. On the other hand, they sometimes do a pretty good job of matching paint at the paint store.
In practice, especially in controlled (e.g., studio) conditions, you can do at least as good a job as the paint store does, if not much better. Dr. Roy Berns at RIT has demonstrated that you can overcome the limitations inherent in a camera's color filter array by making multiple exposures taken with and without carefully-selected Wratten filters and combining the resulting images with some high-powered math. I have some ideas for how to get similar (but probably not as good) results with a much less exotic workflow that I'll be testing out once my order for a couple such filters gets fulfilled in ``six to eight weeks'' from a couple weeks ago. Worst case, I'll be able to replicate his results but with his klunky workflow. I'll report on my findings, of course. Cheers, b&
Cameras are (essentially) linear devices.
In respect to what? ;) We have irregular shapes of spectral characteristics of CFA. Consider what happens when the intensity of red light is too small to cause a reasonably noiseless response in the green and blue channels. -- Best regards, Iliah Borg
And before we know it we'll be back to rotating filter disks :) Edmund On Wed, Jun 5, 2013 at 11:17 PM, Iliah Borg <iliah.i.borg@gmail.com> wrote:
Cameras are (essentially) linear devices.
In respect to what? ;)
We have irregular shapes of spectral characteristics of CFA. Consider what happens when the intensity of red light is too small to cause a reasonably noiseless response in the green and blue channels.
-- Best regards, Iliah Borg
_______________________________________________ Do not post admin requests to the list. They will be ignored. Colorsync-users mailing list (Colorsync-users@lists.apple.com) Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/colorsync-users/edmundronald%40gmail...
This email sent to edmundronald@gmail.com
You jest, but I've actually seriously thought about jury-rigging something like that if not actually seeing if I can scare something up.... b& On Jun 5, 2013, at 2:29 PM, edmund ronald <edmundronald@gmail.com> wrote:
And before we know it we'll be back to rotating filter disks :)
Edmund
On Wed, Jun 5, 2013 at 11:17 PM, Iliah Borg <iliah.i.borg@gmail.com> wrote:
Cameras are (essentially) linear devices.
In respect to what? ;)
We have irregular shapes of spectral characteristics of CFA. Consider what happens when the intensity of red light is too small to cause a reasonably noiseless response in the green and blue channels.
-- Best regards, Iliah Borg
_______________________________________________ Do not post admin requests to the list. They will be ignored. Colorsync-users mailing list (Colorsync-users@lists.apple.com) Help/Unsubscribe/Update your Subscription: https://lists.apple.com/mailman/options/colorsync-users/edmundronald%40gmail...
This email sent to edmundronald@gmail.com
I jest not, it's the obvious way to go. The only problem really is getting filters that fit the lens you are using, maybe a linear strip is the way to do it. Incidentally, I guess if you add enough filters you will have a teleradiometer :) Edmund On Wed, Jun 5, 2013 at 11:35 PM, Ben Goren <ben@trumpetpower.com> wrote:
You jest, but I've actually seriously thought about jury-rigging something like that if not actually seeing if I can scare something up....
b&
On Jun 5, 2013, at 2:29 PM, edmund ronald <edmundronald@gmail.com> wrote:
And before we know it we'll be back to rotating filter disks :)
Edmund
On Wed, Jun 5, 2013 at 11:17 PM, Iliah Borg <iliah.i.borg@gmail.com> wrote:
Cameras are (essentially) linear devices.
In respect to what? ;)
We have irregular shapes of spectral characteristics of CFA. Consider what happens when the intensity of red light is too small to cause a reasonably noiseless response in the green and blue channels.
-- Best regards, Iliah Borg
_______________________________________________ Do not post admin requests to the list. They will be ignored. Colorsync-users mailing list (Colorsync-users@lists.apple.com) Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/colorsync-users/edmundronald%40gmail...
This email sent to edmundronald@gmail.com
On Jun 5, 2013, at 2:53 PM, edmund ronald <edmundronald@gmail.com> wrote:
I jest not, it's the obvious way to go. The only problem really is getting filters that fit the lens you are using, maybe a linear strip is the way to do it. Incidentally, I guess if you add enough filters you will have a teleradiometer :)
Dr. Berns has shown that one or two filters at most (in addition to the unfiltered view) is all you need to, yes, essentially replicate a teleradiometer. See http://art-si.org/ for lots of information. And 4" Wratten (or other) gel filters fit in many standard filter holders, such as those made by Lee or Cokin. The only challenge is how quickly you want to swap them. In the studio, it's not a problem. In the wild.... Cheers, b&
Nice reference. Of course you could have a switchable illuminant setup; eg a set of LEDs or. Edmund On Wed, Jun 5, 2013 at 11:59 PM, Ben Goren <ben@trumpetpower.com> wrote:
On Jun 5, 2013, at 2:53 PM, edmund ronald <edmundronald@gmail.com> wrote:
I jest not, it's the obvious way to go. The only problem really is getting filters that fit the lens you are using, maybe a linear strip is the way to do it. Incidentally, I guess if you add enough filters you will have a teleradiometer :)
Dr. Berns has shown that one or two filters at most (in addition to the unfiltered view) is all you need to, yes, essentially replicate a teleradiometer. See http://art-si.org/ for lots of information.
And 4" Wratten (or other) gel filters fit in many standard filter holders, such as those made by Lee or Cokin. The only challenge is how quickly you want to swap them. In the studio, it's not a problem. In the wild....
Cheers,
b&
Well, this would offer the opportunity to use more than 3 filters => multi-spectral imaging. -Gerhard Am 05.06.2013 23:35, schrieb Ben Goren:
You jest, but I've actually seriously thought about jury-rigging something like that if not actually seeing if I can scare something up....
b&
On Jun 5, 2013, at 2:29 PM, edmund ronald <edmundronald@gmail.com> wrote:
And before we know it we'll be back to rotating filter disks :)
Edmund
At 2:24 PM -0700 6/5/13, Iliah Borg wrote:
Cameras are (essentially) linear devices.
In respect to what? ;)
We have irregular shapes of spectral characteristics of CFA. Consider what happens when the intensity of red light is too small to cause a reasonably noiseless response in the green and blue channels.
Noise and linearity are orthogonal issues as far as I know. Lars
On Jun 5, 2013, at 11:11 PM, Lars Borg wrote:
At 2:24 PM -0700 6/5/13, Iliah Borg wrote:
Cameras are (essentially) linear devices.
In respect to what? ;)
We have irregular shapes of spectral characteristics of CFA. Consider what happens when the intensity of red light is too small to cause a reasonably noiseless response in the green and blue channels.
Noise and linearity are orthogonal issues as far as I know.
It is about how the sensor reacts to low levels, and what type of noise reduction is used on the sensor and in data processing chain before raw is written. All of a sudden channel response to spectrum changes because for low intensity light the lap of the bandwidth is not recorded at the same way. -- Best regards, Iliah Borg
At 1:22 PM -0700 6/5/13, Ben Goren wrote:
Cameras are (essentially) linear devices.
Depends on what you mean by essential. For the cameras I've calibrated the raw values for the CC24 gray steps don't track, even after adjusting for flare etc. This hints at the camera not being linear.
Everything else, especially ACR, instead invisibly translates blackbody radiation sliders into those channel multipliers using some secret sauce recipe.
The recipe is plainly readable in the code of the DNG SDK and follows the DNG specification and a well-known matching procedure as documented in the code. Both are public, hardly a secret. As shown in the code ACR doesn't assume black body radiator for D65.
[T]he camera imposes an inherent error. The colorimetric problem starts there, in the camera. I don't know if a sensor or combination of filters can be made that would make a colorimetric match possible.
Agreed. Every camera is a compromise. Lars
On Jun 5, 2013, at 7:37 PM, Lars Borg <borg@adobe.com> wrote:
At 1:22 PM -0700 6/5/13, Ben Goren wrote:
Cameras are (essentially) linear devices.
Depends on what you mean by essential. For the cameras I've calibrated the raw values for the CC24 gray steps don't track, even after adjusting for flare etc. This hints at the camera not being linear.
Well, I'll admit that I haven't delved into the internals. However, I can take a raw image of a chart, develop it as a UNIWB 1.0 linear gamma TIFF with either dcraw or Raw Photo Processor, extract the chart values from it, and load those values into a spreadsheet and do meaningful linear functions on those numbers and still get meaningful results. And much of my workflow assumes linearity of the original numbers in that raw file. After white balance, and especially after gamma and exposure adjustments? Of course -- it's not even close to linear at that point. But the raw data, at least to some very useful approximation, with all the usual caveats about noise and the like, is close enough to linear that I haven't gotten burned (yet) by pretending that it is. Cheers, b&
At 7:51 PM -0700 6/5/13, Ben Goren wrote:
But the raw data, at least to some very useful approximation, with all the usual caveats about noise and the like, is close enough to linear that I haven't gotten burned (yet) by pretending that it is.
Well, I made that assumption as well. Maybe it depends on what accuracy level you're aiming for. If I remember my numbers, the tracking error for neutrals was in the range 0.5 - 1 dE. The camera noise level was much less than that. I lack standard deviation numbers for the spectral measurements though. (who does?) Lars
On Jun 5, 2013, at 8:07 PM, Lars Borg <borg@adobe.com> wrote:
If I remember my numbers, the tracking error for neutrals was in the range 0.5 - 1 dE
If your neutrals are within 1 DE before profiling, then you should be golden, as any decent LUT profile with a sufficient patch count should have no trouble correcting that. For that matter, I'm personally just fine with 1 DE errors after profiling. In real-world conditions, if you're that close then there's something else (such as the shirt you're wearing) that's likely throwing something off more than that. High single digit DE errors after profiling worry me, but not so much if they're in colors too saturated to reproduce anyway. Double-digit DE errors need to be fixed, especially if average errors are out of the low single digits. I'll also note that you can easily introduce more than single-digit DE errors in linearity just by a lighting setup that's less than perfect that you haven't normalized with a flat-field image. If I found 1 DE errors in linearity, I'd assume that the hardware really is linear for all practicum porpoises and that it's something in my setup that was introducing the non-linarity. Cheers, b&
I did some quick tests some years ago and never saw any non-linearity as long as one stays well below saturation. It is possible that recent cameras have shoulders - on the other hand, CC errors might come from optical crosstalk between channels - light diffusing inside the sensor. Edmund On Thu, Jun 6, 2013 at 1:58 PM, Iliah Borg <iliah.i.borg@gmail.com> wrote:
For the cameras I've calibrated the raw values for the CC24 gray steps don't track, even after adjusting for flare etc.
What was the dark frame and flat field data?
-- Best regards, Iliah Borg
_______________________________________________ Do not post admin requests to the list. They will be ignored. Colorsync-users mailing list (Colorsync-users@lists.apple.com) Help/Unsubscribe/Update your Subscription:
https://lists.apple.com/mailman/options/colorsync-users/edmundronald%40gmail...
This email sent to edmundronald@gmail.com
On Jun 6, 2013, at 10:17 AM, edmund ronald wrote:
I did some quick tests some years ago and never saw any non-linearity as long as one stays well below saturation. It is possible that recent cameras have shoulders - on the other hand, CC errors might come from optical crosstalk between channels - light diffusing inside the sensor.
You are right in more than one way :) The linearity of the sensor should be also discussed in relation to the response to colour. Suppose we have a monochromatic beam of the wavelength that aligns with the maximum of sensitivity striking the sensel at the right angle. We may see linearity of the response in the range of 8 stops. Add a second monochromatic wave to the same beam, with the wavelength that gets 1/32 (minus 5 stops) response. Now if we decrease the intensity of the combined beam 5 stops, the second wavelength gets no response, or the response will be on the non-linear portion, or it will be lost in flare. -- Best regards, Iliah Borg
The last two weeks, I have been implementing rpp 64 into my workflow. Out of rpp 64 the tiff is opened in PS CS5, and from there to printer, Epson pro 9900 What I have been finding, is, with "color" as the film choice, the the image file on the display is dim, or dark, and flat. I go to K64, and the image on the display nearly looks like the original art. I have found working the image with various controls (knobs, or sliders, or whatever they are called) in rpp 64, and can attempt to get the display image to appear like the original. That's with "color" as the film choice. Or, manipulate the tiff of "color" in PS CS5. I will never expect the replication of the original art to look, on the display, or print exactly, as it appear to my eyes! That's a promise I make to my clients, 100%, the representation will not look exactly as the original! I am not using the spectrophotometer. I don't understand everything I read on this list. Some is useful for me. I am not a digital guru, nor a gear head. I know the rest of you understand. I am pleased to be in the 'rabbit hole'. Just delighted. Capture device, lens, sensor, lights, printer, PA271w, are all custom profiled. I do not use canned profiles. My decision to implement rpp 64 into my workflow is based on the information I have read on 'list' and other essays. I thank Iliah Borg and Andrey Tverdokh for the introduction and help. Now, for me, continue to learn and speed up the rpp 64 process in my workflow. David B Miller, Pharm. D. member Millers' Photography L.L.C. dba Spinnaker Photo Imaging Center Bellingham, WA www.spinnakerphotoimagingcenter.com 360 739 2826 On Jun 6, 2013, at 7:58 AM, Iliah Borg <iliah.i.borg@gmail.com> wrote:
participants (7)
-
Ben Goren
-
edmund ronald
-
Gerhard Fuernkranz
-
Henry Davis
-
Iliah Borg
-
Lars Borg
-
Spinnaker Photo Imaging Center