Jump to content

LFI exposed?


jaapv

Recommended Posts

Advertisement (gone after registration)

If the typographical process did not enlarge one file marginally more the other. But I admit, the evidence is on the side of different images now.

 

I must admit that these are two very carefully handled shots but still, one can not avoid even the slightest shift in angles when changing cameras on a tripod ... :)

Link to post
Share on other sites

  • Replies 70
  • Created
  • Last Reply

Top Posters In This Topic

Top Posters In This Topic

Posted Images

Quite true, but then, all modifications within the raw converter are happening within a 16 bit space, so it is quite unlikely that any manipulations would cause a loss of tonal values. The M8 uses 8 bits for storage, but the 8 bit values are just indices into a 14 bit space of tonal values.

 

Michael--I can argue that direction as well. ;) You are correct that the 14-bit space has been mapped into 256 boxes. And the raw converter restores these values with reference to the surrounding pixels, a fact that _tends_ to reconstruct the full original tonal range. (See? I read your article and recognize at least the basics of the subtleties in the M8's compression system. :p )

 

But the fact remains that the 16384 values of the 14-bit space get compressed into only 256 values which span that space. As the diagram on p 56 of LFI 2/2007 shows, that works to our benefit in the low range--but when we get up to the highlights, the M8 doesn't fare as well as other systems.

 

How much total differentiation is enough? Some argue that since burned-out highlights are irretrievably lost, manufacturers should concentrate more on extending the ability to separate data on the high end.

 

Leica made the right choice in the sense that retrieving details in the shadows is quite advantageous. The M8 is impressive, don't misunderstand. But tossing out data at any point is questionable.

 

--HC

Link to post
Share on other sites

But tossing out data at any point is questionable.

 

Interestingly, I've just learned today that the D300 can't do real 14-bit A/D conversion actually, what actually happens is Nikon still uses a 12 bit A/D converter while giving users the option to UPSAMPLE to 14-bit using their EXPEED (wonderful name isn't it? ... makes you wonder what name they'll possibly give to their OWN image sensor. LOL) processor.

 

So Nikon is trying to create data and Leica think that they have more than enough data ... :rolleyes:

Link to post
Share on other sites

Ok, here's the result ... basically, I scanned the images from the LFI magazine and compared them in layers, if they're one shot differently processed, the images would complete collapse with each other and all you'll see is pure darkness.

 

Not so fast there, Simon! ;)

 

Is your scanner glass of optical quality with no striations? Does the scanning head move at precisely the same speed for every scan? Did you place both images precisely in register at the same point on the scanner?

 

In other words, I think there are probably enough variations in the scanning process that if you scanned the same image twice, the two scans would not be identical.

 

In fact, in the copy you show, there are failures of registration, but I don't see any angular differences, where the shadow meets the floor differently or the car covers a slightly different part of the wall.

 

If there are actually two different images, why did the photographer hunker down behind the camera the second time? He already had framing and all he had done was switch out bodies. And if he bent down behind the camera for both shots, how was he able to get into within a couple pixels of the same position both times?

 

Your test is well thought out, but we've already had Jaap complaining that processing one shot twice isn't fair. Now to take two different images from a magazine and scan them just adds 1) the printing process and 2) the scanning process into the mix.

 

Foul, foul! We must be fair to the magazine. Whether we're casting aspersions or exonerating LFI, we must be above reproach!

 

It is one image. One exposure processed twice. That's the only way we could be certain that the ISO is the same, the shutter speed the same, the focus the same. And when I say "the same," I'm thinking "Nicht das gleiche, sondern dasselbe."

 

It must be one image. Taking Dan's recognition a step further: We must ourselves rise above our pettiness and guarantee that our accusations are accurate. We owe it to our bastions! :p

 

--HC

Link to post
Share on other sites

Not so fast there, Simon! ;)

 

Is your scanner glass of optical quality with no striations? Does the scanning head move at precisely the same speed for every scan? Did you place both images precisely in register at the same point on the scanner?

 

Let me tell you the truth, Howard. I've just spent two hours hand drawing those outlines. :D

Link to post
Share on other sites

Nikon cameras use a 10-bit linear storing space, instead of 8-bits, as the M8 does.

 

My point in previous discussions in this forum was why the original 14-bits output is compressed at all. The M8 isn't a fast camera anyway. I never shot at 2 f.p.s. We have bigger and bigger SD cards. SDHC support is more important for M8 users than 8-bits compression (we would get access to 4, 8, 16 GB cards, and even bigger ones are coming!). In any case, if Leica considers a bit of compression is needed, why not use a wider output space instead of the narrow 256 boxes?

 

I would like to have a clear response from Leica (or LFI). Is this possible? Is the camera too slow using 14-bit files? How much? How big are the files? How the 8-bits compression affects the JPG making internal process? Etc.

 

R.

Link to post
Share on other sites

I only have the German edition, and in my mind the text its not misleading. It only says that the upper picture was changed ("umwandelt") with a special software and firmware directly with 16 bits pro channel, whereas the picture below arose ("entstand") like by the serial models with 8 bit.

 

One might think that the withe colour of the car on the upper picture has more nuances than the one on the lower picture. But it is difficult to tell only from comparing pictures printed in a magazine.

Link to post
Share on other sites

...one of the problems I see with conclusions drawn about 16-bit vs. 8-bit RAW files is that people sometimes assume they're equivalent if the final output (from a given set of samples) looks the same. But what seems to be missed very often is that the 16-bit file has the potential for being a much richer source to work with as it is being converted (including, potentially, the "reach" a given camera has in the highlights as well the abruptness which which highlights blow out)......

 

I'm with Sean on this one. Initially I was not convinced by those wanting 16 bit, the images seemed to me to be more than acceptable and I really like the shadow recovery that is possible - but - as I've taken many hundreds of pictures I'm becoming more sensitive to the blown highlights problem. It occurs too often despite the care I take and if 16 bit would, as all agree it could, improve the hightlight rendition without losing the current performance in the shadow areas then it would be something I would very much like. If it slows the camera down it may have to be a selectable option but slower file writing would rarely be a problem for me.

Link to post
Share on other sites

I'm becoming more sensitive to the blown highlights problem. It occurs too often despite the care I take and if 16 bit would, as all agree it could, improve the hightlight rendition without losing the current performance in the shadow areas then it would be something I would very much like.

“If 16 bit would …” – only they wouldn’t. When highlights are blown, they are blown – it is happening in the sensor, not in signal processing. 16 or 14 bits might give you a slightly better rendition of highlights that are not blown, but that’s all.

Link to post
Share on other sites

My point in previous discussions in this forum was why the original 14-bits output is compressed at all.

Most raw formats employ some form of compression, and customers are complaining when there is no compression available. You may accept the storage requirements of huge raw files, but the time it takes to write these files to the card is an issue. It may be less of an issue when one is mostly working in JPEG mode anyway, but with the M8, the preferred format is DNG.

 

In any case, if Leica considers a bit of compression is needed, why not use a wider output space instead of the narrow 256 boxes?

Because 8 bits are one byte, and storing differently sized chunks would slow down the camera again. That’s why some cameras store 16 bits (two bytes) per pixel, even when only 12 bits are actually used. It’s wasteful, but simpler than repackaging the bits.

 

I would like to have a clear response from Leica (or LFI). Is this possible?

I think whatever Leica ever wanted to say concerning this issue has already been published.

 

Is the camera too slow using 14-bit files? How much? How big are the files? How the 8-bits compression affects the JPG making internal process?

That’s a simple one: The files would be twice as big and would take approximately twice as long to be stored. And the compression doesn’t affect JPEGs, as the only compression applied is JPEG compression – the 8 bit compression is for raw files only.

Link to post
Share on other sites

Shifting the topic a little...one of the problems I see with conclusions drawn about 16-bit vs. 8-bit RAW files is that people sometimes assume they're equivalent if the final output (from a given set of samples) looks the same. But what seems to be missed very often is that the 16-bit file has the potential for being a much richer source to work with as it is being converted ....

 

Actually, we have already seen this becoming a problem. Sandy McC's "cornerfix" program for adjusting the vignetting corrections occasionally produces posterized rings when it has too few bits of data available with which to express the correct red levels. That's because he must work with the 8-bit data points saved in the .DNG file, not the original 14 bit information available to the in-camera vignetting routines. His workaround is to transform into a 16-bit space before beginning to correct vignetting.

 

This problem is one more consequence of Leica's having been "shocked, SHOCKED" to discover that IR contamination is magenta-ing the blacks and yellowing the greens in the M8. With limited computational power inside the M8, they seem to be using tables to sneak the vignetting corrections, both luminance and red, inside the compression routine. This has made it difficult to support many lenses, made the corrections somewhat erratic as the (imperfectly known) aperture changes, and denied us the ability to choose how much of each correction we really need. Passing the whole file through as 16bit data would give this back to software correction after the exposure, at some cost in shooting speed and battery life.

 

scott

Link to post
Share on other sites

I only have the German edition, and in my mind the text its not misleading. It only says that the upper picture was changed ("umwandelt") with a special software and firmware directly with 16 bits pro channel, whereas the picture below arose ("entstand") like by the serial models with 8 bit.

 

One might think that the withe colour of the car on the upper picture has more nuances than the one on the lower picture. But it is difficult to tell only from comparing pictures printed in a magazine.

 

I read the German text. It is unclear whether there were two cameras used or just one, the 16-bit file of which was changed to an 8-bit file using the Leica algorithm, by a special program. Now in the last case the " special" software in a computer can be far more powerful than a camera, so it tells us nothing about the comparative performance of a 16 bits M8 vs an 8 bits M8. I'm not saying is is not a legitimate test, but I think it should have been made clear that it was a simulation, instead of the two cameras side by side, as implied in combination with the rest of the article. If it is indeed two cameras, it is much more informative, but in the light of the illustration, there is at least some doubt about that scenario.

Link to post
Share on other sites

“If 16 bit would …” – only they wouldn’t. When highlights are blown, they are blown – it is happening in the sensor, not in signal processing. 16 or 14 bits might give you a slightly better rendition of highlights that are not blown, but that’s all.

 

The gradiation from not blown to blown would be more gradual. That can be important. No one is suggesting the highlights wouldn't blow at some point.

 

The reason Leica accepted this 8-bit system is that they saw no quality loss with it. But they're new to digital systems and I think they misunderstood the advantages of having a more robust RAW file. Comparing two final files doesn't necessarily tell one what he or she needs to know.

 

As for blown highlights in general (which are common with the Zeiss ZM and many of the Leica ASPH lenses) the solution is to use lower contrast lenses and pull back the exposure somewhat.

 

Cheers,

 

Sean

Link to post
Share on other sites

ASandy McC's "cornerfix" program for adjusting the vignetting corrections occasionally produces posterized rings when it has too few bits of data available with which to express the correct red levels.

 

Bingo! And that's exactly the source for the problems KM-25 was talking about as well, though he didn't realize that.

 

Cheers,

 

Sean

Link to post
Share on other sites

jaapv - it pays to read the article accompanying the text - it gives full explanation to all questions you raise and confirms that there is noting misleading at all.

 

Both pictures are according to the article based on the same raw data obtained by a M8 with very early firmware which could store raw data without compression.

 

The conversion of the picture was made by means of a special software which (in a slow and awkward step-by step procedure, it is explained) could change the data into pictures.

 

The special software, thus, made it possible to produce a linear 16 bit version of the picture as well as a 8 bit version as a result of a "Würzelcodierung", ie by drawing the square out of the 14 bit colour information pro channel which is then stored in 8 bits.

 

(There is probably a better and more precise translation of the orignal Gernam text in the English edition of LFI, and especially the square compression is difficult probablyu both to describe in German, to read in German and to translate into English)

Link to post
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...