P: - Add ISO and Exposure Time to the Noise Profile Tag implementation

The information probably comes from the manufacter (I am not 100% sure). Adobe are in charge of the meaning and instructions of the tags, and as of today the noise profile tag will always contain contain junk, independently of the good will of manufactors, because this tag alone simply has no meaning. If Adobe where to add a "noise model exposure" tag, the noise profile could actually make sense, and manufactors would be to blame if there was junk in the tags.

My remark concerns dng instead of tiff, because dng is a widely used format for smartphone raw. I am not an expert on TIFF, but it seems that it did not originally contained any noise profile tag, so the problem does not apply to them, there is not ambiguity in the tag if there is no tag. (https://developer.adobe.com/content/dam/udp/en/open/standards/tiff/TIFF6.pdf)

It really goes against the very principle of metadata : a box where you can read information without ambiguity, and without needing to search further brand specific information on the internet. The exposure time for which the noise profile is given cannot be found anywhere, and manufacters themselves probably don't know the value anymore.

Again, where does this information come from? Is this applied by Adobe, or is it a guess by Adobe, or is it provided by the manufacturer? And what about the data when DNG isn't being used; what do you find then and what do you expect to do with it? 

If Adobe is placing junk in a tag out of nowhere, OK, tell us that and now we have something to 'fix'. 

It really goes against the very principle of metadata : a box where you can read information without ambiguity, and without needing to search further brand specific information on the internet. The exposure time for which the noise profile is given cannot be found anywhere, and manufacters themselves probably don't know the value anymore.

This tag is currently not defined well, as it lacks conventions. I do think that it is Adobe's role is to bring out clear convention for .dng tags, so that the value can be universally read without more context.

Manufacters don't have bad intents when filling the tags inaccurately, but they probably do not know exactly what to put there as it is not crystal clear. This could be adressed by :

- Modifying the equations p58 of the .dng spec to make it more clear that the noise profile changes with exposure

- Adding a note mentionning that the noise profile tag should be given for the current ISO and exposure

Alternatively, if computing the noise model for every exposures and ISO is not possible for manufacters, the spec could be modified witht two alternatives:

- Add a tag called "noise model exposure", where one can read the exposure for which the noise model was computed.

- Add a universal exposure reference, for which all manufacters are expected to compute the noise model.

Furthermore, if this was the cause of the ambiguity, the noise model would be impossible to estimate at all; Yet manufactor, can estimate noise model, and they do that for every ISO. 

Then it is their job to communicate this without any proprietary metadata to all converters, those that work with or without DNG. Do they? 

DNG is just a container for data, like its cousin TIFF. I'm trying to wrap my head around your request with regard to DNG, which can only contain what is provided from the raw. If there is some noise or exposure ambiguity, what is Adobe supposed to do, make stuff up and embed it into this container?

I don't think so, the baseline exposure is never taken into account during noise profiling : the brightness featured in the noise model is considered to be the direct output, scaled between 0 and 1 using the whitebalance and black level. It is not really important at this point to have brightness consistency bewteen 2 cameras.

Furthermore, if this was the cause of the ambiguity, the noise model would be impossible to estimate at all; Yet manufactor, can estimate noise model, and they do that for every ISO. 

2 different cameras with the same settings (ISO and exposure) will not output the same pixel brightness despite capturing the same scene, and this tag gives the value to offset in order for the 2 images to have a comparable brightness.

Agreed, hence the ambiguity and, the outline of why "it is up to interpretation," again no? 

I am not familiar with exposure compensation and have never used the "baseline exposure" tag. From what I understand, the "baseline exposure" is a gain or offset that should be applied digitally to the image during raw processing. 2 different cameras with the same settings (ISO and exposure) will not output the same pixel brightness despite capturing the same scene, and this tag gives the value to offset in order for the 2 images to have a comparable brightness. Please correct me if my interpretation of this tag is wrong.

If not, then I dont think it is related to noise profiling, since noise profiling aims at modeling the noise of the raw unprocessed image for a specific camera. Different camera indded behave differently, which is why having reliable camera tags is so improtant.

I also note that on page 58, the exposure time dependency on the intercept (called Oi) is omitted, which may explain why it is often overlooked when manufacters are tabulating values.

The official specification (that can be found here https://helpx.adobe.com/content/dam/help/en/photoshop/pdf/dng_spec_1_6_0_0.pdf) is not very clear on that point, and it seems that it is up to interpretation for the camera manufacter.

Page 34 starts the outline of why "it is up to interpretation," no? 

The problem really has to do with the .dng format itself, rather than the raw processing. The official specification (that can be found here https://helpx.adobe.com/content/dam/help/en/photoshop/pdf/dng_spec_1_6_0_0.pdf) is not very clear on that point, and it seems that it is up to interpretation for the camera manufacter.

Softwares such as Lightroom can use this noise model, and would work better for a greater range of camera if the tag did take into account the exposure. It would also be very helpful for image processing research.

You are referred to Adobe’s (current) raw processing products or some other and if so, who's?