The effects of compressing to 8 bits

[carstenw]

Advertisement (gone after registration)

Well, if you would post such a DNG, then at least we would have a new datapoint in the endless stream of 8 vs. 16 bit DNG goodness. As it is, this is just yet another thread which adds no truly new information to the mass of existing threads. Btw, it took me just a few minutes to search the forum for the above threads. The search function here is really quite good, and there are many old, interesting threads.

 

I should add that Hans' work of course continues the technical side of the discussion. I mean all the posts since then. An M8 image with a visible difference would take all this work to a new level.

Edited May 6, 2009 by carstenw

[Advertisement]

Hallo,

Look here: Leica M

[jaapv]

I distance myself from the conspiracy thread....

[plasticman]

I distance myself from the conspiracy thread....

 

I fear that someone has hushed you up, Jaap.

Hello guest!
Please register or sign in to view the hidden content.

[jaapv]

No - but I managed to come up with an alternative explanation

Hello guest!
Please register or sign in to view the hidden content.

[plasticman]

Anyway, looking back I think I've seen and read the majority (if not all) of these threads, though it's useful to be reminded of them.

 

The greatest change that's happened since a lot of them were written, is the firmware update that enabled high-capacity SD cards. I no longer see any reason why Leica deny owners of the M8 the ability to save all the image information the camera has captured.

 

Maybe there would be fewer moans about clipped highlights and banding if we finally had access to all those extra tones?

[carstenw]

It's possible. The 8-bit processing likely also reduces battery consumption, and this may be Leica's real reason, combined with buffer space. Anyway, I see no harm in having the option, and would love to have it myself.

 

To temper the feelings, I cannot imagine that Leica has not done exactly the same tests themselves which I would do if the option was suddenly there. In other words, they take photos which in theory push the depth of information hard, and then compare on-screen and printed out. If they can find nothing, then they wouldn't give us the option. However, they can't make us happy that way

Hello guest!
Please register or sign in to view the hidden content.

[luigi bertolotti]

Advertisement (gone after registration)

Jaap, you do have also the DMR, right ? Why don't you post a pair of examples, possibly with "same or next to" lens, same ISO, same (manual) WB setting (same subject of course... you surely can find an apt one), with both the original M8 and DMR DNGs ? Of course, posting them NOT as images here (or, maybe, just for "reference")... the resize/conversion would invalidate all the test... Large files, of course, but maybe you could use your site to make the files downloadable : I think many of us consider you rather trustable to believe you'll not manipulate the DNGs to prove some thesis of your

Hello guest!
Please register or sign in to view the hidden content.

; this question of 16(14) vs. 8 is periodically sprouting out in the forum, and me, and others I think, would like a lot to have some sure practical data to make our own ideas about. Math tells me that the 8 bit conversion into M8 causes some loss by definition... but in which circumstances this is a really VISIBLE issue, I can't say at all.

[Advertisement]

Hallo,

Look here: Leica M

[plasticman]

I cannot imagine that Leica has not done exactly the same tests themselves which I would do if the option was suddenly there. In other words, they take photos which in theory push the depth of information hard, and then compare on-screen and printed out.

 

Yes, as we all know Leica have an excellent track record in color-testing the M8 output...

Hello guest!
Please register or sign in to view the hidden content.

 

But seriously, I'm more than happy with the colors that they deign to allow us.

[jaapv]

Jaap, you do have also the DMR, right ? Why don't you post a pair of examples, possibly with "same or next to" lens, same ISO, same (manual) WB setting (same subject of course... you surely can find an apt one), with both the original M8 and DMR DNGs ? Of course, posting them NOT as images here (or, maybe, just for "reference")... the resize/conversion would invalidate all the test... Large files, of course, but maybe you could use your site to make the files downloadable : I think many of us consider you rather trustable to believe you'll not manipulate the DNGs to prove some thesis of your

Hello guest!
Please register or sign in to view the hidden content.

; this question of 16(14) vs. 8 is periodically sprouting out in the forum, and me, and others I think, would like a lot to have some sure practical data to make our own ideas about. Math tells me that the 8 bit conversion into M8 causes some loss by definition... but in which circumstances this is a really VISIBLE issue, I can't say at all.

 

I did that once and was flamed for "not the same lens" etc..

My problem is that I have no comparable lenses for the DMR and the M8. The only ones I might compare are either the 28/2.8R and the Summicron 35 asph, or that same Summicron with the 35-70/4.0R, obviously with apertures made the same. Take your pick, I can easily put the full DNGs on Yousendit.

[carstenw]

28/2.8 (latest, I presume) + 35 Cron ASPH, please

Hello guest!
Please register or sign in to view the hidden content.

[jaapv]

Yes. It will be Friday evening I guess.

[carstenw]

I guess you will choose the type of subject matter where you can see the difference between M8 and DMR, right?

Hello guest!
Please register or sign in to view the hidden content.

[jaapv]

I do not have any baby bottoms handy, but I'll try... A sunset might be a good idea? And a wide open landscape?

[plasticman]

I do not have any baby bottoms handy, but I'll try...

 

Hello guest!
Please register or sign in to view the hidden content.

[luigi bertolotti]

I did that once and was flamed for "not the same lens" etc..

My problem is that I have no comparable lenses for the DMR and the M8. The only ones I might compare are either the 28/2.8R and the Summicron 35 asph, or that same Summicron with the 35-70/4.0R, obviously with apertures made the same. Take your pick, I can easily put the full DNGs on Yousendit.

 

Well... I'd say the lens difference is not such a critical issue... after all is the COLORS that "worry" us (and the "separation" between them, not a possible microdifference in lens' dominant) , not micro sharpness or so... I think Cron 35 and 35-70 at 35, both for instance at 5,6 could be a very good test. Using the same focal is really important, I think....is much better to have the same view at the same distance to confront colors "transitions". A landscape with finely graduated colors (sea, meadows) can be ideal... but surely, as other have said, skin is also a very good test.

 

Haven't you a MR adapter ? Just to ask you a harder job... one could focus the 35/70 on the R, exchange R with M8 on the tripod (verify position of the focus plane !), transfer the lens, make some subtle focus bracketing...

Hello guest!
Please register or sign in to view the hidden content.

Edited May 6, 2009 by luigi bertolotti

[Advertisement]

Hallo,

Look here: Leica M

[carstenw]

For the sake of agreement, I will go with Luigi's wish, i.e. 35-70/4 and 35/2A both at f/5.6 or f/8.

 

Jaap, are you *asking* for what kinds of scenes you can see the difference between the DMR and the M8? This is a long-standing claim of yours, so you should *know*. Ahem.

[plasticman]

A landscape with finely graduated colors (sea, meadows) can be ideal...

 

Luigi - these are exactly the sort of subjects that would not exhibit the problem in any definite way. A meadow is a patchwork of micro-contrasts and an array of color variations, likewise a seascape.

[thrid]

So true!

Hello guest!
Please register or sign in to view the hidden content.

 

Anyways, last night I pulled out the back-number of LFI where they discuss the compression algorithm when it was revealed, and frankly even though the article does its best to put a positive gloss on the whole thing I came away much less convinced than before.

Anyone who feels nothing (or next to nothing) is being lost in the process need only read the article and look at the graphs - which are far more eloquent in uncovering the shortcomings of the system than the author probably intended.

 

I agree. Just look at the file size. It shrinks from 20MB to 10MB. That is a massive jump. I may not be a programer, but even I know that you can't fit 10 pounds of mud in a 5 pound bag, without getting some of it all over you... There has to be some loss of data and I don't believe it is just noise, as Leica claims.

 

I still think the smartest and simplest thing Leica could do, is to allow the photographer to decide if they want their files to be compressed or not. Other companies certainly allow that option (Nikon etc).

Edited May 6, 2009 by thrid

[jaapv]

Well... I'd say the lens difference is not such a critical issue... after all is the COLORS that "worry" us (and the "separation" between them, not a possible microdifference in lens' dominant) , not micro sharpness or so... I think Cron 35 and 35-70 at 35, both for instance at 5,6 could be a very good test. Using the same focal is really important, I think....is much better to have the same view at the same distance to confront colors "transitions". A landscape with finely graduated colors (sea, meadows) can be ideal... but surely, as other have said, skin is also a very good test.

 

Haven't you a MR adapter ? Just to ask you a harder job... one could focus the 35/70 on the R, exchange R with M8 on the tripod (verify position of the focus plane !), transfer the lens, make some subtle focus bracketing...

Hello guest!
Please register or sign in to view the hidden content.

 

Duh! I had forgotten about that thing!

[luigi bertolotti]

Luigi - these are exactly the sort of subjects that would not exhibit the problem in any definite way. A meadow is a patchwork of micro-contrasts and an array of color variations, likewise a seascape.

 

Right, sorry ... I thought the same AFTER making my post...

Hello guest!
Please register or sign in to view the hidden content.

; probably the best can be... human skin, skies, even some pleated dress, fabric... Edited May 6, 2009 by luigi bertolotti