unadjusted pixel values

How do I get rawtherapee to show unadjusted raw pixel values?

According to the documentation, even the neutral profile applies the cameras white balance to the displayed picture, and this is born out by showing the raw histogram which shows the red significantly brighter than blue or green, yet the color values shown under the pointer show nearly the same value for each color.

Background: A am attempting to evaluate a light source and camera combination for digitizing film negatives. The light source is color adjustable, and negatives have an orange cast to them that I would like to remove by adjusting the light source to cancel the orange cast. I may also like to bias the color of the light source to cancel the sensitivity bias of the sensor (camera seems much more sensitive to red than green or blue).
I am very new to rawthereapee, so I don’t yet know how to do much with it. The first thing I need to figure out is how to capture film negatives in a way that preserves as much information as possible from the negative. I can worry about converting the result into a positive once I have that nailed down.
For those really into the details of what I am doing, I’m using a Sony aRII with FE 90mm F2.8 Macro G OSS lens. The light source is a Yongnuo digital YN600 RGB with a custom light diffuser. I have built a film carrier, and a table that mounts to a copy stand so I can repeatably place the film between the light source and the camera.

Apply the neutral processing profile, set the demosaicing method to “none”, then zoom in to 100% or more.

A sample file would help.

If you’re not going to create an input profile to match that adjusted light source, then it’s probably better if you go with standard illuminant D50/D65.

“Sensitive” here is ambiguous. Cameras with a Bayer filter, such as the SONY ILCE-7RM2, have the highest quality signal in the green channel, not only due to the nature of the silicon the sensors are constructed from which has the highest spectral sensitivity to green wavelengths, but more so because there are twice as many green-filtered photosites as red or blue.

Since it’s a LED device I would prioritize profiling it once everything is adjusted.

1 Like

(Hmm, I can’t seem to figure out how to quote on this site.)

I applied the neutral processing profile, set the demosaicing method to none and zoomed to 800% where I could clearly see the individual pixels, and see that there are twice as many green pixels as red or blue.

On a picture that is a very short exposure of a uniform white light source, so it is extremely under exposed, I’m seeing numbers like 48% for red, 103.1% for green, and 63.9% for blue.

This does not correlate at all to the raw histogram which shows all three should be less that 25%.

I would be happy if I could get numbers where 100% was full scale, or the raw integer number returned by the pixel. At the moment, I don’t know what I am getting in the navigator display. The values I’m getting don’t make any sense.

I’ve taken a 79 of sample images of three different subjects at multiple exposures. Subject #1 is the white light source at maximum brightness of all five LED colors (happens to be white). Subject #2 is the same light source with an unexposed frame of film negative, and subject #3 is an exposed film negative.
That would be a rather large amount of data to upload. If you can explain how this would be helpful, I might be able to choose one or two of those to upload.
So far I have only been looking at the 1st image which is subject #1 at the shortest exposure.

I think creating an input profile is probably the right way to go. I think that I may need to create an input profile for each brand of film negative I plan to digitize. I expect I have several thousand images to digitize, so I will need to automate the process as much as possible. I have not seen any built in profiles to handle converting film negatives into positives.

What I mean by “sensitive” here is that with a white light source, the green sensor produces a larger value than the blue sensor, and the blue sensor produces a larger value than the red sensor. So in that sense, the green sensor is more sensitive than the blue or the red. These sensors are linear, and since the eye is logrithmic, you get much finer brightness steps at the high end of the sensors range than at the low end. My first attempt is going to be to expose the negative as much as possible without clipping any of the color sensors. But I plan to do some experiments with this to make sure that is the right thing to do. I plan to bias the light source that all three sensors are at about 90% of full scale when capturing an unexposed film negative.

“Since it’s a LED device I would prioritize profiling it once everything is adjusted.”
I’m not sure what you mean here. What should be adjusted before profiling the light source?

@Morgan_Hardwood
I may be on thin ice here, but since he studies the histogram as well as certain pixel values, wouldn’t the state of the histogram button be of interest, as well? I.e. the one which ties the histogram to the gamma corrected output profile or to the working profile?

Have fun!
Claes in Lund, Sweden

Do you have a color target? We need daylight + StdA shots of a color target for the SONY ILCE-7RM2. Instructions here:
http://rawpedia.rawtherapee.com/How_to_create_DCP_color_profiles#Shooting_the_Color_Target

When examining raw values using demosaicing=none in RawTherapee 5.4, click on the RGB column in the Navigator to switch from meaningless % to raw values. RawTherapee 5.5 (and current development versions) will do that automatically.

To create an input profile which takes the camera + light source + film stock + scene light into account, you would need to photograph a color target using the same film as your slides are on, in the same light at the same time . Since the photographs are already taken and the film slides already developed, it’s too late for that, so I don’t know how you could create a color target which takes the film stock into account. The next-best thing is to exclude the scene light and film stock and create a profile which takes “just” your camera + light source into account. You would only need to do this once, assuming you don’t adjust your YN600’s color between shots.

I would be curious to experiment what happens if you hold a translucent, unexposed and developed film slide over the lens, and photograph a color target through it, then use that to create an input profile for those slides backlit using that light at that color using that camera. I will be testing something similar for UV light through a yellow filter. Bogus results expected, but curious nonetheless.

The color of the light source for instance.

It will not affect the raw histogram nor the raw pixel values.

I did this, and the ranges should be 0 to 255, but I’m getting numbers like 271 so these are the same number as the % numbers, just multiplied by a constant. I am still not getting the unmolested values from the camera.
I don’t have time to do more this morning.

No, the ranges are 0 to white level, so numbers like 156860 are to be expected.

When I click on the RGB it cycles between [%], [0-1], and [0-255]. I would assume that these indicate the range of values. In each case when I put the cursor over certain pixels I get numbers larger than the maximum of the range.

Perhaps we are talking about different versions of the software. I am running the latest stable version. Version 5.4.

edit: I opened a file that was taken with a much longer exposure, and I am getting values much larger than 255. I was fooled by the 1st image having max values near 255. It seems the [0-1] range is dividing the value by 255, and [%] is dividing by 2.55. Neither is very useful, they should be dividing by the full scale value instead.

In any case it looks like I am getting unscaled raw value in the [0-255] mode, so it looks like I can proceed with my testing. Thanks

1 Like

We should describe how to examine raw values in RawPedia, and describe what those values mean (i.e. are they before or after black level subtraction, etc).