Paul Kay
Member
I have taken many HDR images on my Canons taking multiple exposures and using them to create high bit depth files which are then remapped and output as 8-bit files. As has been stated, the results can be spectacularly good.
However I also use the same software (Tone Mapping from Photomatix) on Leica M8 files which I process as '16-bit' conversions from Raw. Whilst I fully accept that theses files are not true 16-bit files and that theoretically at least I must be losing information, in PRACTICE the M8 files and Tone Mapping work extremely well together and the results can be, although different from true HDR generated images, extremely effective.
Attached is an example taken on an M8 with 24mm Elmarit and IR filter at 160ISO (ie very 'straight').
This is not 'magic' but does work AND I cannot achieve the same results direct from a 12-bit Canon file, which is because of DIFFERENCES in the files. As Richard says, if you haven't actually sat down with M8 files, its difficult to appreciate their strengths (and weaknesses) but this doesn't mean that they don't exist and are merely a figment of Leica afficianado's hyped imaginations!
However I also use the same software (Tone Mapping from Photomatix) on Leica M8 files which I process as '16-bit' conversions from Raw. Whilst I fully accept that theses files are not true 16-bit files and that theoretically at least I must be losing information, in PRACTICE the M8 files and Tone Mapping work extremely well together and the results can be, although different from true HDR generated images, extremely effective.
Attached is an example taken on an M8 with 24mm Elmarit and IR filter at 160ISO (ie very 'straight').
This is not 'magic' but does work AND I cannot achieve the same results direct from a 12-bit Canon file, which is because of DIFFERENCES in the files. As Richard says, if you haven't actually sat down with M8 files, its difficult to appreciate their strengths (and weaknesses) but this doesn't mean that they don't exist and are merely a figment of Leica afficianado's hyped imaginations!
Attachments
xihalife
Member
Oh well... fortunately Leica knows the camera is not perfect and is making improvements to it. If they actually listened to some opinions on this forum they would probably cancel their plans to upgrade the M8 hardware.
I don't doubt that M8 is an excellent camera. I have taken many great shots with it. But the fact is that the lack of precision bites me sometimes. One very challenging shooting environment is the winter in Finland, where we have sun reflecting from snow and ice (extremely bright) in contrast with trees and other objects that are not covered by snow.
Yesterday I was out shooting and I used both M8 and a Canon DSLR. Unfortunately some of the M8 photos turned out to be useless due to the fact that there is no room for brightness overflow. The light meter is useless in such conditions as it is fooled by the difficult conditions and the LCD screen is not bright enough for bright sunlight, so I had to try and use a best guess.
With M8, you absolutely need to make sure the brightest point in the scene stays within the limits of the brightness ... 1.0 or less. This tends to mean the rest of the image is very dark. While the M8 is quite good at dark colors, the more extreme you go the more noise and less color detail you tend to get in the dark areas. I will post some examples later on. I *was* able to get good shots.
I use the M8 a lot so I don't really know why mr. Marks keeps saying I am not qualified to state an opinion until I've tried it. The mentioned HDR trick does nothing to salvage lost highlights and it does nothing to improve color that is lost in dark areas. It only works if the shot is ok to begin with, which is to say that the colors lie in the range that the 8-bit file can store them.
I'm totally sure it works ok if the light conditions are reasonable (again - that is where the "high" comes from the HDR). A single shot in 14 bits of resolution gives you more dynamic range than in 8 bits. I don't doubt that if the shot is perfect and the color detail is not lost, the HDR conversion works and the HDR tools can be used for nice effects.
sitemistic - I'm realizing this. And in fact it makes no difference to state my opinion here. Anyone who believes M8 is perfect and needs no improvement is already happy with their camera and will not agree with my wishes of getting improved functionality. What bugs me most with this 8-bit issue is that it *seems* it could be changed in software. There's a lot that would need hardware upgrades (such as an improved sensor to bringht lower base ISO and less noise) but this seems like something that could be done a firmware update.
I don't doubt that M8 is an excellent camera. I have taken many great shots with it. But the fact is that the lack of precision bites me sometimes. One very challenging shooting environment is the winter in Finland, where we have sun reflecting from snow and ice (extremely bright) in contrast with trees and other objects that are not covered by snow.
Yesterday I was out shooting and I used both M8 and a Canon DSLR. Unfortunately some of the M8 photos turned out to be useless due to the fact that there is no room for brightness overflow. The light meter is useless in such conditions as it is fooled by the difficult conditions and the LCD screen is not bright enough for bright sunlight, so I had to try and use a best guess.
With M8, you absolutely need to make sure the brightest point in the scene stays within the limits of the brightness ... 1.0 or less. This tends to mean the rest of the image is very dark. While the M8 is quite good at dark colors, the more extreme you go the more noise and less color detail you tend to get in the dark areas. I will post some examples later on. I *was* able to get good shots.
I use the M8 a lot so I don't really know why mr. Marks keeps saying I am not qualified to state an opinion until I've tried it. The mentioned HDR trick does nothing to salvage lost highlights and it does nothing to improve color that is lost in dark areas. It only works if the shot is ok to begin with, which is to say that the colors lie in the range that the 8-bit file can store them.
I'm totally sure it works ok if the light conditions are reasonable (again - that is where the "high" comes from the HDR). A single shot in 14 bits of resolution gives you more dynamic range than in 8 bits. I don't doubt that if the shot is perfect and the color detail is not lost, the HDR conversion works and the HDR tools can be used for nice effects.
sitemistic - I'm realizing this. And in fact it makes no difference to state my opinion here. Anyone who believes M8 is perfect and needs no improvement is already happy with their camera and will not agree with my wishes of getting improved functionality. What bugs me most with this 8-bit issue is that it *seems* it could be changed in software. There's a lot that would need hardware upgrades (such as an improved sensor to bringht lower base ISO and less noise) but this seems like something that could be done a firmware update.
Paul Kay
Member
xihalife said:
xihalife
If we go back to the original post the poster seemed to be shocked at finding the M8 to supply only 8-bit files. I think that the response has been (certainly from myself) that provided that you work within the camera's existing limitations, it is capable of very effective results. The 8-bit vs. 16-bit issue can be highly hypothetical and as R Marks put it 'philosophising' is all very well but is no substitute for trying something for yourself.
I really do not believe the M8 to be perfect but it does have really useful attributes and yields somewhat different files to those from many other digital cameras. I could go on at length about the less pleasant attributes of my M8 or indeed my Canons (still can't get good saturated reds from them for example) but I prefer to work within the existing limitations of a camera and learn its strengths and weaknesses.
What I for one have tried to say in this thread is not that the M8 is perfect or even that it is better than other cameras. I have tried to show that the camera is capable of results which might prove surprising given that it supplies 8-bit files. Whether or not someone wants to use it should depend not on whether its specification is 8 or 16-bit, but whether they are prepared to accept and work within its current limitations - which might not be as dramatic as many people seem to think that those created by bit depth issue actually are.
xihalife
Member
sitemistic said:
Well, this is a good guess, right? We don't really know what the truth is. Leica has officially stated that they experimented with 16-bit RAW files and decided that there was no visible difference. If true, it would mean that the bits are there. I'm just not convinced they tried all the possibilities to be sure that the extra bits are never useful.
Even if that was true, the image processor in the M8 is not very powerful so it would be interesting to save the non-corrected RAW photo and do the vignetting correction in RAW import step. Who knows, sometimes you might want to do a "LOMO" style photo and just leave the corners as-is!
xihalife
Member
Paul Kay said:
Fair enough. However I am quite confident Leica will address this issue in one of the future upgrades... possibly after they upgrade the sensor.
And sure - I am a reasonably happy M8 owner and I keep using it with its limitations and trying to push the limits as much as I can. Accepting the fact that the RAW images contain less dynamic range than some other cameras is important. When you realize it, and accept it, you will also accept the fact that you simply need to work around some issues. Saying that there is absolutely no difference between 8 and 16 bits causes more confusion... I think it's quite important to face the fact that the dynamic range is not there so you need to be more careful with lighting when you take photos.
jaapv
RFF Sponsoring Member.
xihalife said:
Too simplistic I fear. this issue has been chewed to death. No, 8 bits is not equal to 8 bits, strange as it may seem. There was an excellent article in LFI on the issue. DNG allows different ways of recording the depth, which means that the M8 behaves more like a 12 to 14 bit camera.
xihalife
Member
jaapv said:
I've seen an article which describes a use of a look up table to compress 14-bit values. The trouble is that the look up table is indexed using 8-bit values, which means there are 256 pieces of 14-bit values. This is still not enough to capture a wide dynamic range, it's only enough to describe those 256 values more accurately. And no, the M8 does not behave like other cameras which save 12 or 14 bit data... that is a known fact. The lost highlights are one proof of that - M8 is the only high end camera I have seen which seems to clamp all color values to 1.0.
Keep in mind Leica have said that they did experiment with saving of full bit depth so we know more bits could be stored. Some M8 users feel JPG is enough and RAW is useless. Good for them. I would like to use the full detail, thank you. It should be an option in the menu.
dcsang
Canadian & Not A Dentist
ya know.. xihalife, I think I've learned more about bit depth and how it impacts all cameras by reading some of your posts in this thread (and the articles posted by Hub and the one I found myself).
I really think you understand why this is a bit of an issue to some - definitely not to those that don't care about it or to those that are perfectly happy with the M8 as it is currently.
Ever since the advent of RAW in digital cameras, that's what I've been using for the exact same reason you stated - I want as much information as I can gather - to have a camera that's been crippled by a manufacturer merely because they feel that the photographer really doesn't "need" the information is a bit presumptuous on the manufacturers part - regardless of who that manufacturer is...
Dave
I really think you understand why this is a bit of an issue to some - definitely not to those that don't care about it or to those that are perfectly happy with the M8 as it is currently.
Ever since the advent of RAW in digital cameras, that's what I've been using for the exact same reason you stated - I want as much information as I can gather - to have a camera that's been crippled by a manufacturer merely because they feel that the photographer really doesn't "need" the information is a bit presumptuous on the manufacturers part - regardless of who that manufacturer is...
Dave
jaapv
RFF Sponsoring Member.
There were interesting graphs, comparing. I was convinced...xihalife said:
I've just thought of something. I'll shoot the same scene with an M8 and the DMR side by side and post the RAW files. That way everybody can see what is happening between 16 and 8 bits.
It may take one or two weeks, I'm short of time....
Last edited:
xihalife
Member
jaapv said:
I will need to find out the article you are referring to. The one I saw linked to in this discussion earlier was basically pointing out how the compression causes artifacts, which are not usually visible in the resulting image, but will stand out if the colors are adjusted in post processing. This mirrors my own experience so I found it rather credible. (but again - I guess I'm repeating myself - most of the time I'm very happy with the images produced by the M8).
ferider
Veteran
Look, in a perfect world, the perfect M8 would have menus to pick the lens, output raw format, and an external software package to correct for vignetting/cyan shift, etc. on the user's PC/MAC. No lens upgrades/coding necessary, etc.
In the real world, however, where much marketing is done over the internet, a brand is created by amateur lens testers who immediately and _only_ would use raw formats, and then complain about vignetting and necessary color correction, a company that makes much revenue via lens sales and support (lens upgrades, etc) has no choice. The picture coming out of the camera has to be as perfect as possible, even if a few bits color depth are lost. If, in the process, some revenue can be made via lens upgrades, and the shooting bandwidth of the camera is improved, even better. Also, who shoots color with Leicas anyways - the M8 is the perfect B+W digital rangefinder .... (brand reference).
I predict that a full frame Leica will have the same issues. A new, FF sensor with deeper colors will be used (say 24 bits ?), and some, this time more, since vignetting is stronger, bit depth will be lost in the camera. Still the user will get deeper colors on output (12 bit or so ?). We'll see ...
Remember:
In the real world, however, where much marketing is done over the internet, a brand is created by amateur lens testers who immediately and _only_ would use raw formats, and then complain about vignetting and necessary color correction, a company that makes much revenue via lens sales and support (lens upgrades, etc) has no choice. The picture coming out of the camera has to be as perfect as possible, even if a few bits color depth are lost. If, in the process, some revenue can be made via lens upgrades, and the shooting bandwidth of the camera is improved, even better. Also, who shoots color with Leicas anyways - the M8 is the perfect B+W digital rangefinder ....
(brand reference).I predict that a full frame Leica will have the same issues. A new, FF sensor with deeper colors will be used (say 24 bits ?), and some, this time more, since vignetting is stronger, bit depth will be lost in the camera. Still the user will get deeper colors on output (12 bit or so ?). We'll see ...
Remember:
- 1 stop vignetting = loss of 1 bit color depth.
- Higher ISO = more bits per pixel
Last edited:
hub
Crazy French
Richard Marks said:
Not files. The camera. As Dave was saying, it is hard to find a camera to try out because nobody around seems to have one, nobody rents one, and there is no way I'd shell out 5 Large + a lens just to try it.
If you argue that I haven't tried it, don't just offer to provide files. That's not trying out a camera.
hub
Crazy French
xihalife said:
1. the number of bits does not preclude the dynamic range in any way. It is just an indication of the precision of the information. The more bits you have the more precision you get, ie real difference between to value very close, that might be rounded up. See it has the number of decimal digits (round to the nearset cent vs rounding to the nearest 5 cents). Also this include the noise, which the downside.
2. if you readuce 14-bits value to 8-bits like it is done, you LOSE infromation. in the end you have lost that precision and you can't recover it. Note that for those who say "it is because they don't have enough power" I'd reply that they actually require more processing to perform this operation.
And FYI, some lower end Nikon do that, but the quantization table provide an everage of 9.4bits of information.
See http://www.majid.info/mylos/weblog/2004/05/02-1.html
That lead to another question: is the JPEG file generate on camera by these 8bits RAW data samples or is it done from the upstream data? But who can answer as the source code of the fimrware is not available, nor any documentation.
ferider
Veteran
hub said:
Formally speaking that is wrong. The number of bits does define the dynamic range for digital signal processing.
Much like dB in the analog context.
----------
Added: I guess I should elaborate - sorry for the geek talk:
Assume you have a black-box processing an input signal and generating an output signal. The "Dynamic Range" of the box is defined by ratio of largest signal divided by smallest signal that you can transmit. Which means, if you add, on input, a small signal to a large signal you should still see a different output than if you just transmit the large signal. For a digital box, the smallest signal is 1 bit. The largest 111111... (repeat nr. of bits you can transmit). The dynamic range is proportional to log(11111...../1), i.e. proportional to the nr. of bits.
It's a well defined term in engineering.
Last edited:
Richard Marks
Rexel
xihalife said:
Forgive me! You can and of course express any opinion you like and this is much appreciated. It is simply that it is a bit more believable if it is qualified by factual experience.
I think we shall have to agree to differ.
A most informative thread for all concerened.
Best wishes
Richard
sandymc
Member
Without wanting to spoil the fun, the 8-bit to 14-bit comparason's been done, at least in synthetic form:
http://www.l-camera-forum.com/leica-forum/leica-m8-forum/33768-8-bit-versus-14-bit-dngs.html
Sandy
http://www.l-camera-forum.com/leica-forum/leica-m8-forum/33768-8-bit-versus-14-bit-dngs.html
Sandy
sandymc
Member
ferider said:
Sorry, not so. Number of bits defines dynamic range only for systems using linear encoding. Which the M8 definitely is not.
Sandy
Paul Kay
Member
I guess that the bottom line problem is actually determining how all this theory actually translates into practice. On the one hand you have people like myself, and clearly others too,who have a practical view and have actually tried the M8 (and other cameras) and have come to the conclusion that the M8's 8-bit files are perfectly adequate if utilised to shoot subject matter in ways that suits both us and the camera. On the other hand there are theorisers and users to whom the issue is a significant drawback and who either believe that they need the highest bit depth possible or find aspects of their shooting style and subject matter is not suited to Leica's 8-bit output.
Not sure where the discussion can go to from here if I'm totally honest!
Not sure where the discussion can go to from here if I'm totally honest!
ferider
Veteran
sandymc said:
Not sure what you mean by linear. Linear in terms of Gallois field function ?
Do you know if or not the M8 uses a non-linear code ? I doubt it.
In any case, the encoding does not matter. It's the information quantity that matters. Measured in bits. Go the other way and you will see. Take an 8 bit signal, and encode into, say 12 bits. You still have 8bit dynamic range.
Last edited:
jaapv
RFF Sponsoring Member.
Hmmm. Your posts read that you are having problems with the files you can expect. I would say it was a fair offer.hub said: