RAW/TIFF Printing

[Colin Southern]

You got what I had expected to get.

What I am getting is what doesn't make any sense.

Best guess is you've got a setting different somewhere. I did mine in CS5 so it's not using the latest process - not sure if that would make any difference.

At the end of the day, DNG is still a RAW format -- it just has the data saved in a standardised structure. Off memory (going back a few years now) v1.1.0.0 used to drop black masked pixels (which - arguably - could be used for noise filtering calculations), but that was resolved in v1.1.0.1 - after that I'm not aware of any data being dropped until the newer "compact" version options came along.

If it helps, I can send you my files so you can see if it's the DNG or RAW version that's changing (convert it yourself and compare it to my conversion).

[Mark von Kanel]

What both of you are getting makes no sense to me..... but maybe next year when ive changed from a porky engineer in his 40's to a beach body Greek Adonis and some one who knows how to use photoshop rather than getting a headache as soon as i open it....

[Colin Southern]

What both of you are getting makes no sense to me..... but maybe next year when ive changed from a porky engineer in his 40's to a beach body Greek Adonis and some one who knows how to use photoshop rather than getting a headache as soon as i open it....

Essentially, what Manfred is doing is opening up 2 "versions" of the same image, and then comparing them (ie telling the program to only display the differences - and since there shouldn't be any, the result should be a black square). The 2 "versions" are the one straight from the camera and one that's supposed to be identical that's been converted to DNG from the same raw file straight from the camera.

[Manfred M]

Best guess is you've got a setting different somewhere. I did mine in CS5 so it's not using the latest process - not sure if that would make any difference.

At the end of the day, DNG is still a RAW format -- it just has the data saved in a standardised structure. Off memory (going back a few years now) v1.1.0.0 used to drop black masked pixels (which - arguably - could be used for noise filtering calculations), but that was resolved in v1.1.0.1 - after that I'm not aware of any data being dropped until the newer "compact" version options came along.

If it helps, I can send you my files so you can see if it's the DNG or RAW version that's changing (convert it yourself and compare it to my conversion).

Would you mind sending me your files Colin? I went through the excercise again with a different set of files yesterday evening and got identical results.

I have found that the workflow / method used tends to change the results:

When I use the "Place" command, it puts the new layer in as a Smart Object and I get the result shown (which I can replicate with other images). My other DSLR is in Portugal right now with my daughter, so I could not see if this was something specific to the D800 or not.

I find that the method that I use does give different results:

When I open the NEF and the DNG as separate files and then do a "select all" and then do a copy and paste, the resultand image in the expected pure black. I suspect I might have come across an "undocumented feature" a.k.a. a bug in CS6.

[Manfred M]

What both of you are getting makes no sense to me..... but maybe next year when ive changed from a porky engineer in his 40's to a beach body Greek Adonis and some one who knows how to use photoshop rather than getting a headache as soon as i open it....

From one engineer in his 50's to one in his 40's, Colin's explanation is right on, but I thought I might provide a slightly more technical detail (just to try to give an an even bigger headache).

That one really important feature in Photoshop is the ability to do things in layers. I simple analogy is that this feature is somewhat like working on a different transparent film and then stacking them on top of each other. This feature allows the user make non-destructive changes to the piece being worked on. Unlike a physical layer of material, there are some very powerful functions that can be applied to these layers.

My understanding of a DNG file is that it is really just a "wrapper" for a raw file, and it encapsulates the raw data out of the camera, the external sidecar (xmp) file as well as a thumbnail of the image. This means that the actual image file that is reconstructed from either file format should be identical. To put this into practice, I imported the same file from my camera in two different ways, both times using the import capabilities in Adobe Bridge. Bridge will let you import a RAW file as is or gives you the option of converting it to a DNG file; so I did both.

My next step was to bring both versions into Photoshop CS6, and storing them in two different layers. Photoshop allows the user to perform various operations on these layers. As data is just data, there is a feature called blending modes that allows the user to change how these two layers interact. In "normal" mode, the data in the top layer will block the view of the data lower down in the layer stack, just like in the transparent film scenario. Any areas that do not have data, i.e. transparent areas) allow the layers below to show through, which is great for photo compositing.

There are also a number of other blending modes that allow mathematical manipulation of the data between the top layer and the layers below. The “subtract” and “difference” blending modes subtract the values in the same position on the two layers (the difference in the two modes is how they handle underflow values; i.e. if subtracting the values results in a value that is less than zero). One would expect that if both layers contain identical data and we subtract the two values, we should get a result of “0”, i.e. total black.

This is not what I ended up getting, so one would conclude that something happened to during the DNG conversion that means that the two files are not identical. This would normally indicate that there are data changes during the conversion to a DNG file; for instance compression artifacts from lossy compression. There is nothing in the Adobe literature to indicate that they apply lossy compression, so I assume they are compressing using lossless algorithms. This suggests that there is a software bug causing this. As I noted in a note to Colin; this occurs with a specific workflow (using Photoshop’s “Place” functionality). I did check to ensure that this was not a pixel misalignment issue. This does not occur when I when I do a copy & paste method; in that case I get the expected totally black image.

[Mark von Kanel]

Thanks guys, clear as crystal... (reaches for paracetamol)

[Colin Southern]

Would you mind sending me your files Colin? I went through the excercise again with a different set of files yesterday evening and got identical results.

I have found that the workflow / method used tends to change the results:

When I use the "Place" command, it puts the new layer in as a Smart Object and I get the result shown (which I can replicate with other images). My other DSLR is in Portugal right now with my daughter, so I could not see if this was something specific to the D800 or not.

I find that the method that I use does give different results:

When I open the NEF and the DNG as separate files and then do a "select all" and then do a copy and paste, the resultand image in the expected pure black. I suspect I might have come across an "undocumented feature" a.k.a. a bug in CS6.

Yep - sure. I'll send you a link when I get into work.

I just opened both in PS and used the move tool to drop on onto the other as a stacked layer, with snap turned on.

[Colin Southern]

My understanding of a DNG file is that it is really just a "wrapper" for a raw file, and it encapsulates the raw data out of the camera, the external sidecar (xmp) file as well as a thumbnail of the image.

Not really a "wrapper" - it really just restructures the existing metadata into a standardised and publically documented format. Sidecar files don't come into it because they're only generated as a result of changes to a RAW file (so a DNG file supports the subsequent data that WOULD HAVE been written to a sidecar file if the same changes were made in a DNG file as they were to a RAW file, but a DNG conversion doesn't take into account any sidecar file that may already exist).

There IS also an option to fully encapsulate an original RAW file into a DNG for "purity and light archival considerations", but I doubt many bother using it as all it really achieves is a roughly doubling of the size of the file.

[Colin Southern]

Would you mind sending me your files Colin?

Here ya go ...

https://docs.google.com/file/d/0B2gx...it?usp=sharing

[Manfred M]

Here ya go ...

https://docs.google.com/file/d/0B2gx...it?usp=sharing

Thanks Colin - Interesting. With your files I got exactly what I expected, pure black. I followed the same workflow that I used for my images that gave the unexpected results.

My D90 arrived home from Portugal tonight (along with my wife and daughter), so I am going to see what happens with files out of that camera. I'm wondering if this is a Nikon issue or just a D800 anomaly.

Time enough tomorrow for further investigations.

[Colin Southern]

Thanks Colin - Interesting. With your files I got exactly what I expected, pure black. I followed the same workflow that I used for my images that gave the unexpected results.

My D90 arrived home from Portugal tonight (along with my wife and daughter), so I am going to see what happens with files out of that camera. I'm wondering if this is a Nikon issue or just a D800 anomaly.

Time enough tomorrow for further investigations.

Happy to test your files if you want to fire them back my way.

[Manfred M]

Not really; no printer can reproduce the quantity of data that is contained in an image file. At best, a printer is restricted to a few hundred thousand distinct colours, while the data itself (and many screens) can handle millions of colours; depending on the colour space in use.

[rpcrowe]

I will agree that most online printing sources will only accept JPEG files. Additionally, our local COSTCO membership stores have pretty comprehensive photo departments which only accept JPEG images.

OTOH, there are two local brick and mortar custom printers that I sometimes use who will accept either TIFF or PSD images as well as JPEG. However, being small quantity custom printing businesses; their prices are quite a bit higher than are most online sources.

I have not made a comparison of prints done from the custom printers (using PSD or TIFF) against the output from Internet printing sources (using JPEG). However, I suspect (but have no documentation to prove) that there would be a greater differential in the quality of prints from various Internet JPEG printing sources (at greatly differing prices) than in the difference between JPEG and TIFF input.

I have one image which I would like printed for wall display on a very large stretched canvas. The range of prices for this service is pretty broad. I would expect better quality from mid-range to the higher price printers than from the absolute cheapest sources. Bottom line is that IMO you usually get the quality you pay for.

This is of course predicated on the fact that I would supply the best quality image possible for printing.

[revi]

Have I seen that phrase "print as tiff vb.net" before (in the post from getpass), with an associated link to rasteredge.com?

Sounds like another spam to me, tbh.:
- restarting an old thread
- question is non-sense as posted (no such thing as 'tiff vb.net format')
- and poster has only just joined, and only the one post to his name (joined today, last activity shortly after time of post)

edit: indeed, the mention of "tiff vb.net" has appeared in similar form (and with the exact same link to rasteredge.com) at least twice before in posts by readimage and gdjump, and both these users also joined the day they posted, only have one post and have never reconnected since joining and posting.

[Dave Humphries]

Thanks for the detective work Remco; getpass, gdjump and readimage now banned.