What do we lose from RAW to TIF

[jeremyrh]

Probably a dumb question but...

I was wondering what we would lose if we open a RAW file in one software and don't apply any processes at all such as tone curves or white balance but then output a TIF file to a second software.  If the demosaic steps are equivalent in the 2 softwares, is there anything lost in doing this?

The reason for asking was that I was wondering about the more general question of if I lose anything by exchanging TIF files between a number of softwares, e.g. RAW conversion in DxO, coloutr balance in C1, sharpening in Topaz, or whatever.

[PeterAit]

I am not sure your question makes sense--the part about demosaicing. It's only done once--a TIF is already demosaiced. If you need to brush up on the process, Wikipedia has a good article.

Here's how I would test. Open the raw file in the software of interest and export a tif--16 bits of course. Then open both the raw and the tif in a program, such as LR, that allows side-by-side comparison. You can now compare color, dynamic range, etc to see what, if anything, has been lost.

[jeremyrh]

I am not sure your question makes sense--the part about demosaicing. It's only done once--a TIF is already demosaiced. If you need to brush up on the process, Wikipedia has a good article.

Here's how I would test. Open the raw file in the software of interest and export a tif--16 bits of course. Then open both the raw and the tif in a program, such as LR, that allows side-by-side comparison. You can now compare color, dynamic range, etc to see what, if anything, has been lost.

The point about demosaicing is just that if one software does it better than another then this may introduce a bias in the outcome.

Your test makes sense but i don't trust myself to evaluate the results. Also - maybe changes (rounding errors or whatever) don't show up immediately but only at the end of a chain of processes.  (In my day job (seismic data processing) we typically perform calcualtions in 32bits but evaluate the results on an 8bit display for similar reasons.)

[kimballistic]

Any algorithm in your favorite raw processor that optionally reaches into the raw data to improve its operation will not have that option when working on a tiff.

White balancing is the most well-known example.  You can use white balance controls on tiffs/jpgs but they are vastly inferior to white balancing a raw (which as I understand it happens before the data is demosaiced!).

I don't personally know of other examples but I bet there are some forum users who do.

[DP]

White balancing is the most well-known example. 

you can output to "tiff" w/o WB - it all depends on your raw converter...  for example RPP can output w/o applying such adjustments.

[DP]

Probably a dumb question but...

I was wondering what we would lose if we open a RAW file in one software and don't apply any processes at all such as tone curves or white balance but then output a TIF file to a second software.  If the demosaic steps are equivalent in the 2 softwares, is there anything lost in doing this?

it all depends on what you do before and after demosaick steps... for example some software does operations like WB before demosaicking (as per-channel multiplication) and some does after (a more complicated math like in ACR using data from DCP camera profiles) ... so not only demosaicking shall be identical, but the whole pipeline shall be aligned if you want the eq. results... I think DxO can output linear DNG files to make sure there is no WB applied to the data yet - no ?

[digitaldog]

Raw is not rendered and you have to render it or it looks damn ugly:

The entire idea and a major goal of photography is to render the image (or print) as you, the image creator desires. It's why many of us capture raw instead of a JPEG which is a rendering produced by a machine.
This superb article is must read for those considering their roles rendering images:
http://www.lumita.com/site_media/work/whitepapers/files/pscs3_rendering_image.pdf

[gjbloch]

Excellent article. Thanks for sharing.

[dgberg]

A very good read.

[xpatUSA]

This superb article is must read for those considering their roles rendering images:
http://www.lumita.com/site_media/work/whitepapers/files/pscs3_rendering_image.pdf

I was going to mumble something about input-referred vs. output-referred but that article says it all about that and more!

Thank you for the link.

[rdonson]

Thanks, Andrew!!!

[kirkt]

Here are a couple of articles that might be of interest:

https://trumpetpower.com/photos/Exposure.html

and

http://translate.google.com/translate?u=http%3A%2F%2Fwww.guillermoluijk.com%2Farticle%2Fsuperhdr%2Findex.htm&langpair=es%7Cen&hl=EN&ie=UTF-8

Kirk

[digitaldog]

The question should (?) or can be: What do we gain from raw to TIFF.

[xpatUSA]

The question should (?) or can be: What do we gain from raw to TIFF.

With my no-demosaicing-needed Sigma, I gain a bit of color:



Ignore EXIF - FastStone Viewer strikes again

[Alan Klein]

Would Windows Bitmap BMP or Paint create a file without changes?

[digitaldog]

With my no-demosaicing-needed Sigma, I gain a bit of color:

Now if it could only aid in focus 

[DP]

Now if it could only aid in focus 

actually you don't want targets in perfect focus, I let you try to think why

[digitaldog]

actually you don't want targets in perfect focus, I let you try to think why

Go ahead and make something up.

[dreed]

If you go back to the Canon 1DS Mark I, the "raw" output mode produced .TIF files. The contents of Canon's CR2 files are still roughly aligned with the TIFF structure but there's now special bits in there for Canon and renaming them doesn't make them "just work". That doesn't mean TIFF files are the same as raw files, just that they're both big for similar reasons