Jump to content

LFI exposed?


jaapv

Recommended Posts

  • Replies 70
  • Created
  • Last Reply

Top Posters In This Topic

Top Posters In This Topic

Posted Images

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.

But that’s not really a workaround but just the way this should be done. The 8-bit values are mere indices into the 14-bit space of tonal values, and all manipulations of the image data should properly be applied within this 14-bit space (or 16-bit space, 16 bit being the closest available word size).

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.

Point taken, you are right. However there is a noticeable lack of "control" in the highlight areas when attempting to edit pictures in CS3. This contrasts markedly with the "control" available in the shadow areas. You don't have to attempt much in the way of changing the highlight rendition to run into posterization, banding and other problems. I would prefer to have an option to improve this aspect even if it slowed the camera down and created large files.

 

Leica are probably absolutely correct when they say that as far as the printed images are concerned there is no discernable difference between a "16 Bit" image and an "8 Bit" compressed / decompressed image taken straight from the camera. But that is not the point. Hardly any picture I want to print can't be "improved" by the judicious use of the controls available in programs like CS3. The ability to make those adjustments, which were routine in the darkroom, is an essential aspect of using an M8 and it appears to have been to some degree compromised.

 

It is my perception that the number of pixels which contain no information is increased by the "8 Bit" files, thus making blown highlights physically larger. This, if I'm correct, reduces detail and resolution. Some blown pixels are almost inevitable but there just seem to be too many.

Link to post
Share on other sites

But that’s not really a workaround but just the way this should be done. The 8-bit values are mere indices into the 14-bit space of tonal values, and all manipulations of the image data should properly be applied within this 14-bit space (or 16-bit space, 16 bit being the closest available word size).

 

Just for clarity, CornerFix has always, from the earliest versions, done its corrections in a 14-bit linear space; what happens is that the data is first decompressed from 8-bits to 14-bits, then corrected.

 

The difference between the early versions of CornerFix and the current "16-bit" versions is that the old versions then recompressed the data to 8-bits. This was done to produce a file that was entirely indistinguishable from an original M8 file. The current versions however leave the data at 14-bits unless you enable the "Level Compression" option. Getting rid of the recompression step is what gets rid of the "concentric circles"......The only downside to this is that some raw developers, e.g., C1, assume the data to be 8-bit, and so need the level compression enabled.

 

Sandy

Link to post
Share on other sites

There WAS a prototype M8 that did carry 16 bits in very early firmware and leica decided visually it made no difference between 8 bit and 16 bit.

 

Yes, and that was a mistaken method for judging, IMO. Its based on a somewhat restricted understanding of why higher bit files can be useful.

 

Cheers,

 

Sean

Link to post
Share on other sites

Hardly any picture I want to print can't be "improved" by the judicious use of the controls available in programs like CS3. The ability to make those adjustments, which were routine in the darkroom, is an essential aspect of using an M8 and it appears to have been to some degree compromised.

 

I hope people come to understand that point. Many professional working with M8 files, myself included, have seen that they can sometimes be problematic in the highlights. A true 12 - 16 bit file would likely provide more highlight flexibility. It would be less abrupt.

 

Again, this is related to the issue KM-25 raised when he discussed the reaction of some magazines' art staff to highlights in M8 files. It doesn't help, in this respect, that the camera has access to some of the highest contrast lenses ever made.

 

Scott, Sandy, Peter and myself are all looking at various aspects of the same issue.

 

Cheers,

 

Sean

Link to post
Share on other sites

...Many .. with M8 files.... have seen that they can sometimes be problematic in the highlights.....

 

It is compounded when the centre weighting auto exposure cannot recognise the picture highlights which are going to blow out, or if the meticulously examined in-camera histogram [i always use a magnifying loupe] says that the highlight wall is comfortably protected, yet the image opens with blown highlights.

 

Last weekend , needing to work fast and making my first intensive 'people' [as opposed to landscape] images my M8 drove me to exasperation; auto-exposure, even with a 2/3 stop compensation could not cope in several shots without blowing out highlights and the clunkier manual metering turned out to be equally tricky given the problem I described with the unreliability of the in-camera histogram. Unsatisfactorily, I ended up needing to dump several files which otherwise would have been kept..

 

There is much about the M8 that is fabulous, but I am beginning to think that by far it's most significant weakness is caused by the sequence of design decisions which can lead to weak and blown highlights. In my darkroom days I hated those contrasty prints by some printers who presented enlarging paper-white as a substitute for highlight detail, the equivalent in digital prints can appear worse.

 

The mantra 'expose to the right' has unravelled with my M8 with consequences for my files, and how fast I can work whilst producing exemplary exposures. I love the camera, but hate how I have to manage highlight protection with it.

 

.............. Chris

Link to post
Share on other sites

Chris--

Remember that because of the M8's file compression, exposing to the right is less advantageous with it than with other cameras.

 

Remember also that even though the manual doesn't mention it, the effective metering pattern of the M8 varies with different lenses. With wide-angles, it works more like a center-weighted averaging meter, but with longer lenses, the pattern tightens to function more like a spot.

 

--HC

Link to post
Share on other sites

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.

 

The meaning of the original question is how 8-bit compression of the original 14-bit data (to a smaller 8-bit output file) affects the internal JPG making process, in terms of time. I don't know for sure, but making a JPG from a 14-bit raw output could be much slower than making it from 8-bits raw output.

Link to post
Share on other sites

There WAS a prototype M8 that did carry 16 bits in very early firmware and leica decided visually it made no difference between 8 bit and 16 bit.

 

Was the camera much slower with early firmwares? It depends on the SD card used?

 

How much slower would be the camera using a wider output space (non linear 10-bits or 12-bits or uncompressed 16-bits)?

 

We know the effects, positive or negative, of this compression, but we don't know the magnitude of those effects. How big are the losses in highlight detail? How much slower would be the camera working in a wider output space (uncompressed 16-bits or compressed 10, 12 bits, etc.?

Link to post
Share on other sites

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.

It might be quite a bit more than twice as long - the camera would have to apply all vignetting corrections to 16-bits files as well, how long that would take with the limited computing capability af a digital camera is anybodies guess.

Link to post
Share on other sites

.... because of the M8's file compression, exposing to the right is less advantageous with it than with other cameras...

 

Howard - Thank you, I understand the advantages/disadvantages of M8 shadows'highlights, but in the problem exposures I encountered the post-shot camera histogram appeared perfectly safe; the histogram was telling me that my highlights were stopping well short of the highlight wall. Upon opening, the highlights were in fact blown.

 

I don't quite see how any measuring method could compensate for specular highlights....

 

Jaap - I can't tell if your statement was prompted by my post. If it was I agree with you, but the issue of blown highlights in my problem exposures was not specular.

 

................. Chris

Link to post
Share on other sites

Howard - Thank you, I understand the advantages/disadvantages of M8 shadows'highlights, but in the problem exposures I encountered the post-shot camera histogram appeared perfectly safe; the histogram was telling me that my highlights were stopping well short of the highlight wall. Upon opening, the highlights were in fact blown.

 

 

 

Jaap - I can't tell if your statement was prompted by my post. If it was I agree with you, but the issue of blown highlights in my problem exposures was not specular.

 

................. Chris

I see, Chris :( In that case I would direct you to the Bibble website. I posted this in another thread just now.

Link to post
Share on other sites

........We know the effects, positive or negative, of this compression, but we don't know the magnitude of those effects. How big are the losses in highlight detail?

 

I guess that with a "16 Bit" file one might reasonably assume that the “loss” in the highlights would then be no more than the “loss” in the shadows. As it is we have significantly less control in the highlights. So if nothing else with a “16 Bit” file we would have the ability to manipulate and modify all parts of the image in the same way and to the same extent as we currently can with the shadow areas.

 

I for one think this is a worthwhile objective though I fully accept that “16 Bit”, or some other comparable format, would have to be a selectable option as for a lot of people it would not make sense. My objective is to get the highest possible print quality out of my Leica lenses. I’m now more and more convinced that this is currently being made virtually impossible in a lot of situations because of the ”8 Bit” file structure.

Link to post
Share on other sites

The meaning of the original question is how 8-bit compression of the original 14-bit data (to a smaller 8-bit output file) affects the internal JPG making process, in terms of time. I don't know for sure, but making a JPG from a 14-bit raw output could be much slower than making it from 8-bits raw output.

Not at all, because the 8-bit data couldn’t even be used to generate JPEGs – the 8-bit values are just indices into a table holding the actual 14-bit data. So if JPEGs had to be created from compressed 8-bit raw data, they would have to expanded to 14 (or rather 16) bits first, which would be even slower – and quite pointless.

Link to post
Share on other sites

Was the camera much slower with early firmwares? It depends on the SD card used?

It would obviously depend on the SD card, but the camera with the pre-production firmware couldn’t be used for reliable measurements. It’s not like there was an early M8 writing 16-bit DNG files like the DMR does, or at least I am not aware of one. The 16-bit output of this camera more closely resembled a memory dump – no EXIF metadata or anything, just the raw data. This “raw format” was as raw as it gets. I suppose that there will always be a firmware version like this at some point during the development of a camera: The camera can take photographs at last, and dump the raw data onto the card, so the engineers have actual samples to play with when devising the image processing algorithms to be implemented in the final firmware.

Link to post
Share on other sites

I guess that with a "16 Bit" file one might reasonably assume that the “loss” in the highlights would then be no more than the “loss” in the shadows. As it is we have significantly less control in the highlights. So if nothing else with a “16 Bit” file we would have the ability to manipulate and modify all parts of the image in the same way and to the same extent as we currently can with the shadow areas.

On the contrary, all raw file formats, Leica’s compressed 8-bit format and Nikon’s lossy NEF format included, give much more control over the rendition of highlights than over that of the shadows. If you divide the entire dynamic range into 1 EV sized chunks, than an uncompressed format will waste one half of all the available “buckets” on the highest EV value, 1/4 on the second highest, 1/8 on the third and so on. In this trickle-down-economy of raw formats, adding more bits will make the already rich (the highlights) much richer, but achieve very little for the poor (the shadows). Still, with linear sensors and linear A/D converters, there isn’t anything else one could do.

 

For example, Canon went from the 12-bit format of the EOS-1Ds Mark II to 14-bit in this year’s EOS-1Ds Mark III. This means that the highest 1 EV chunk will be sliced into 8192 parts rather than the 2048 parts the EOS-1Ds Mark II provided – that’s twice as many as the 4096 “buckets” the Mark II used for its entire dyamic range! Now did they make the move to 14 bits because 2048 steps hadn’t been sufficient for rendering a mere 1 EV? Of course not. They did this because you have to throw a lot of bits at the sensor data just to get a couple of additional buckets for rendering shadow detail.

 

Leica, on the other hand, never believed 12 bits to be sufficient, which is why they opted for a 14-bit format for the M8, ensuring that the rendering of shadow detail would be superior to that of the (then current) 12-bit DSLRs. However, it is in the nature of linear raw data formats that most of the values are essentially wasted – you don’t really need thousands of buckets for resolving one EV worth of highlights, but that’s what you get if you digitize the sensor data with a resolution fine enough to get good shadow detail. So Leica chose a compressed format reducing the ridiculously high resolution of highlights, but at the same time preserving the resolution of shadow detail. The result is a 8-bit format requiring less storage space than a 12-bit format, but giving superior results in the critical shadow areas, similar to a 14-bit format.

 

So in the raw files created by the M8, there is still much more highlight than shadow detail, only the difference isn’t quite as pronounced as with other raw formats. An uncompressed 16-bit format would give you even more resolution within the highlights, but you’ve already got plenty. The real issue isn’t highlight detail within the sensor’s dynamic range, but blown highlights. A 16-bit format would give you still better resolution of the highlights that aren’t blown, but you don’t need that. When highlights are blown, the sensor pixels have reached saturation; still more light wouldn’t lead to more electrons being accumulated. This clipping occurs within the analog sensor, and neither the A/D converter nor any subsequent digital processing could change that. The only remedy for blown highlights is to correct for the over-exposure.

Link to post
Share on other sites

It might be quite a bit more than twice as long - the camera would have to apply all vignetting corrections to 16-bits files as well, how long that would take with the limited computing capability af a digital camera is anybodies guess.

 

That's something I've quietly been thinking about all year. The M8 works best with coded lenses so that it can do its internal magic on the files before writing the DNGs. It would take more computing power, for sure, to do all those corrections at a higher bit level. I think that when Leica first realized the IR problem, they must have thanked their lucky stars they had both coding and 8-bit RAW files in place.

 

Cheers,

 

Sean

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...