It seems to me compressed DNG is better than uncompressed?

[kennylovrin]

Hey guys

When I got my M9 I read a bit about compressed vs. uncompressed DNG on the M9 and it seemed that the general consensus was that the difference is so small that it can barely been seen, but that a lot of people shoots uncompressed because it has more data. Of course, a file that is twice the size must contain more data, that isn't really debatable in that sense. Hence I've been shooting uncompressed so far without thinking much about it.

However, I never found a good comparison and now I'll be going on a two week trip to Thailand next weekend, and I want to simplify what I bring along which led me to investigate this because it makes a lot of difference when the files I get are 50% smaller than now. So this is not a discussion wether I should bring my laptop or not, I will not bring it and that is the end of it. I will however buy some extra SD cards and I'll be able to dump some shots onto my iPad.

Anyway, I set up a tripod today and shot four images; two properly exposed where one was uncompressed and the other compressed DNG; two like the first two but underexposed 2 stops; two versions of the underexposed shots, pushed 2 stops in Aperture.

All of this was done in Aperture, and I've spent a good hour now looking at the shots in pairs at 100% (I wasn't enjoying it) and to be honest, I can't really tell them apart, and if I HAD to choose one as the better, I consistently chose the compressed file (without seeing at the time which was which).

I am slightly surprised over this, but with RAW files I guess it can differ from software to software etc. Perhaps Aperture just manage to handle the compressed version better than the uncompressed, what do I know.

The question is though, did someone compare this themselves, and what was the conclusion? I am not looking for an answer like "uncompressed has more data so I shoot that" because that is a fact. What I am interested in is if someone ACTUALLY can see a difference. I mean completely honestly, see a difference that isn't purely psychological. It seems Leica claims there is no quality loss, but on the other hand it is impossible to remove data and not lose information, so if there was in fact no loss, even theoretically, there would be no point in supplying the two modes.

Also, does anyone know exactly what the camera does to produce these compressed files from the uncompressed ones?

One thing I find interesting as well is that the uncompressed files from the M9 are about 10mb larger than they were from my 5dmkii. And the M9 has lower pixel count, so the question there is if everyone else supplies RAW files that are equivalent to what Leica refers to as compressed, as they make more sense comparing the pixel count and file size against the 5dmkii for example? Because it doesn't make sense that "RAW" data is 10mb larger from a sensor with less pixels to me. Unless the bitdepth is different? So is the uncompressed mode something Leica has, that no one else give the end user access to in practice?

Thanks!

Kenny

 

[theno23]

I've not tried it myself, but I switched back and forth a couple of times and can't say I noticed much difference in quality.

However, I have a bit of an information theory background and I can tell you that it's possible for two files, one with more data than the other to hold the same amount of information. That's what ZIP-type compression does.

I've never heard a clear explanation of what kind of compression Leica uses in M9 DNG files, though there might be one out there. I think DNG gives you more than one option.

- Steve

 

[EdwardKaraa]

Keeny, would it be possible to post a couple of photos? It would be interesting if we would see what you see 🙂

 

[kennylovrin]

I've not tried it myself, but I switched back and forth a couple of times and can't say I noticed much difference in quality.

However, I have a bit of an information theory background and I can tell you that it's possible for two files, one with more data than the other to hold the same amount of information. That's what ZIP-type compression does.

I've never heard a clear explanation of what kind of compression Leica uses in M9 DNG files, though there might be one out there. I think DNG gives you more than one option.

- Steve

True, it could be a lossless compression, but I am not sure if it is, or if it is in fact throwing information away? I guess that leads back to the question of "why have two options" if the compressed DNG still is decompressed by the RAW engine. Unless there is a performance hit for the computer to handle the compressed version (didn't notice that though).

 

[kennylovrin]

Keeny, would it be possible to post a couple of photos? It would be interesting if we would see what you see 🙂

Sure thing! I will post the original DNG as I would expect that is what people what to fiddle around with themselves. However, to se what I personally saw, TIFF or PNG exports would probably be more accurate, but if someone is interested in that specifically I can provide that as well (for example if someone isn't using Aperture).

Try this link! https://www.dropbox.com/sh/etgo1333ccydtz0/ReSlpbtK-F
Obviously the smaller files are the compressed ones! Also, the third version I did of each was just taking the two latter shots and pushing the exposure +2 to see what happened.

Actually, perhaps I should share my train of thought here:

I saw one comparison somewhere online where it was concluded that the uncompressed shots were ever so slightly sharper than the compressed ones. I couldn't really see that difference in the examples, it was so small that it could just be a side effect of the RAW demosaic. What I suspected though was that the compression would convert from 14bit to 12bit or something like that, effectively throwing away some fidelity in the levels, which is why I did the specific test of underexposing and pushing it back up - I thought it would show in the shadows as it was most likely the range to be affected by a lesser bit depth.

Anyway, I can't really say anything like that happened. 🙂

I'm sure that someone might see something I'm not seeing, but so far I think it is pretty safe to say that the difference is so small that you have to make it up to see it more or less. 🙂

BUT! Of course, a print might show some difference in gradation, or a better screen than mine (2009 iMac). It could be that the compressed files throw away info that aren't really visible on a screen, but would be in print (then again, I would suspect a bigger difference when post processing compressed files then).

 

[willie_901]

Kenny,

Destroying information during lossy compression is not a disadvantage if that information is unneeded, unusable or redundant.

Retail printers and most computer displays can not use all the data in an uncompressed tiff rendered from an uncompressed raw file. So using a jpeg for these is not an issue.

It seems the scene you photographed did not require the all the raw information. Perhaps you should repeat this experiment with a scene where the dynamic range is a challenge.

I prefer to use losses compression and I convert my uncompressed raw images (Nikon or Fuji) to DNGs with Lightroom.

I doubt there is a disadvantage to using compressed DNGs on your trip. Simplifying your storage card handling seems worthwhile.

 

[hlockwood]

BUT! Of course, a print might show some difference in gradation, or a better screen than mine (2009 iMac). It could be that the compressed files throw away info that aren't really visible on a screen, but would be in print (then again, I would suspect a bigger difference when post processing compressed files then).

There you go, Kenny. I don't think looking at a screen is the proper determinant. Try making a good sized print for the more accurate comparison.

Harry