M8 Review at B&H

[ho_co]

Advertisement (gone after registration)

Georg--

Yes, 16-bit does preserve data better than 8-bit, as you say.

 

Michael Hußmann's article in LFI is at pains to point out that in the M8, the full 14-bit range of the M8 sensor is maintained, but stored in a smaller space.

 

That is, the range of values up to 2^14 is still restored in processing the DNG, so the 16-bit TIFF does still make sense. (Some intermediate values are lost, but will generally be recovered in the pixel interpolation process.)

 

Your question is a good one, but remember, this answer is being by a person whose understanding of the math involved is less than stellar.

 

--HC

[AlanG]

So far as I know, humans can't tell the difference between 8-bit and n-bit images (n>8) on screen or in prints. In fact, virtually all printers only use 8-bit data. However, and this is a big however, if you want to post-process an image using (e.g.) curves, color balancing, saturation changes etc. and if you only have 8-bits to work with you are at risk to end up with posterization in tonality and/or hue etc.... and the rest that you wrote.

 

 

Yes, that is exactly how I see it too. If the raw processor only has 8 bits of information to work with, I don't see how it can get out any more than 256 shades from red, green, and blue regardless of how it is encoded. And if the camera has a wide dynamic range, then these shades will really be spread out on files that make use of this range. So it would seem to me that any 8 bit raw file would have to restrict one's ability to adjusted the image in post processing compared with a raw file that stores a higher bit depth of data.

 

 

And ho_co wrote "That is, the range of values up to 2^14 is still restored in processing the DNG, so the 16-bit TIFF does still make sense. (Some intermediate values are lost, but will generally be recovered in the pixel interpolation process.)"

 

I really don't really understand how the raw processors such as C-1 can extract this information if it really exists. But for all I know, maybe it can. I wonder why Canon is moving to a higher bit depth if one can somehow encode 16 bits of data into 8 bits of space and not lose anything significant. Doesn't Canon know how to do this too?

 

A discussion of mathematics is fine but I am only interested in the results. And I know that digital photography is full of all kinds of interpolation. I'd really like to judge it for myself. I'm open to examine the subject and have my opinion changed if someone can provide comparative "matching" raw files shot with an M8 and another fine camera that records a higher bit depth.

[Advertisement]

The Leica Q3 is available for pre-order at B&H Photo Video and Adorama

[J_Brittenson]

Let me ask a hopefully not to stupid question. When I develop DNG files in C1 into TIF files, I have the choice beteen 8 and 16 bit. Given that the DNG file has (only) 8 bit, does the 16 bit option still make sense? (I know that when I make adjustments under 16 bit I loose less data than if I make them under 8 bit.)

It makes good sense to use 16 bits. Demosaicing, noise reduction, and many other operations will extend the bit depth. The most naive debayering for instance, that of averaging each sample with its 8 neighbors, will increase depth by 3 bits. Algorithms that operate only on chroma but leave luminance (detail) alone will give less, but you still get 1.5-2 bits of depth during demosaic and noise reduction. Coarse luminance noise reduction can be done by tweaking the threshold where demosaicing stops at luminance edges. Then the image can be polished later in a second stage. Both of these effectively increase bit depth since they act as low-pass filters.

[J_Brittenson]

So far as I know, humans can't tell the difference between 8-bit and n-bit images (n>8) on screen or in prints.

This is often repeated in online forums, and it's grossly wrong. Just about everyone can make out a doubling in luminance (hence the choice of gamma 2 for NTSC and its prevalence in RGB), so the number of bits required is log2 of the luminance range of the output device. A flat screen with a 2000:1 contrast (luminance range) for instance will require 11 bits for smooth gradation.

 

The other problem is that RGB is really a format to drive CRT guns, with one value for each primary, so it has the same inherent gamut limitations as CRT screens. The luma of the most saturated primary for instance (say RGB 255,0,0 for fully saturated red) is only one third the luma of white (RGB 255,255,255) since only one gun is firing. The same goes for the other primaries obviously. The fully saturated secondaries (say RGB 255,0,255 for magenta) are roughly twice as bright as the primaries give or take some variance in eye color sensitivity. This results in a not so practical, contorted, gamut shape. To fit a good range of hues at a good range of luminance values requires an RGB gamut to be huge.

[ho_co]

And ho_co wrote "That is, the range of values up to 2^14 is still restored in processing the DNG, so the 16-bit TIFF does still make sense. (Some intermediate values are lost, but will generally be recovered in the pixel interpolation process.)"

 

I really don't really understand how the raw processors such as C-1 can extract this information if it really exists.

My quote, to be sure, Alan, but Hußmann's point: the raw processor doesn't so much 'extract' as 're-create' the data. In all digital cameras which use a Bayer-pattern sensor, 50% of all green values are interpolated, as are 75% of red and blue values. The M8 uses this 'automatic' feature to fill them in.

 

A discussion of mathematics is fine but I am only interested in the results.

Agreed. I think that's why we're all so impressed with the incredible image quality of the M8. Recall David Adamson's earlier thread on that topic.

 

Adamson's comments also would indicate a basis for a manufacturer which isn't scoring so well to consider revising matters, perhaps by increasing bit depth. That is the simple, reasonable next step.

 

Remember as well that Nikon's compressed Raw stores more than twice as many values as the M8, but I haven't seen anyone say that the NEFs are better.

 

Leica has been very sophisticated and very open in the matter, the rascally bastards!

 

--HC

[leicapfile]

Boy it sure is nice not to have to pull out a calculator to evaluate an image/camera.

 

Just enjoy the images no matter what you use to capture them. After all, photography is a VISUAL media.

 

Jerry

[Concorde-SST]

Advertisement (gone after registration)

No i do not think so .... you are just playing with some very dangerous lightroom sliders here..... Ligtroom has several sliders to be carefull with .. very easy to process a file to death and completely ruin it!

Pull the red slider to the right for instance and see how it messes up your B&W file if you zoom in and look .......

 

Yeah thats exactly what I mean. I can "mess" around with the sliders as

much as I want with Canon 1Ds MkII files, and the file is spotlessly clean.

This difference I don´t understand.

 

best,

 

Concorde-SST

[sdai]

An ideal test to show maximum quality from each camera would be for the Canon to be set at ISO 100 and the M8 to be set at ISO 160.

 

Alan, I've just pulled out all the back issues of Color Foto in the past two years and found this ... anybody who are ready to beat these lab test numbers, now it's show time.

 

1Ds2 Resolution when tested with a lowly 50/2.5 macro:

 

ISO 100: 1594 lp/ph

ISO 400: 1596 lp/ph

 

Note ... the 1Ds2 actually resolves more at higher ISO ...

 

DMR with a 100 APO Macro:

 

ISO 100: 1313 lp/ph

ISO 400: 1301 lp/ph

 

M8 with 50 lux asph.

 

ISO 160: 1304 lp/ph

ISO 320: 1253 lp/ph

[AlanG]

Alan, I've just pulled out all the back issues of Color Foto in the past two years and found this ... anybody who are ready to beat these lab test numbers, now it's show time.

 

lp/ph

 

That's interesting stuff, but what I wanted to see has nothing to do with resolution. It is the dynamic range and the ability to adjust the files. (Especially if the M8 is only working in 8 bit.)

[AlanG]

Yeah thats exactly what I mean. I can "mess" around with the sliders as

much as I want with Canon 1Ds MkII files, and the file is spotlessly clean.

This difference I don´t understand.

 

best,

 

Concorde-SST

 

So you have both cameras and could shoot a controlled test with them. (I've outlined what I'd like to see in a test.) If it isn't too inconvenient, please do so and give us access to the raw files so we can decide for ourselves how the 8 bit encoding on the M8 compares with 12 bit of the 1DsII.

[awolf]

I just posted some pictures on Zenfolio | awolf. Misc_07, Misc_06 and Carmel_06 are taken with a P20 Phase One back on a Rollei 6008 (16-Bit?). The M8_07 is, neutrally, with the M8. As much as I like some aspects of the M8, it is not, unfortunately,a match to the color quality and dynamic range of the P20. Whatever it is, the bit #s, the sensor size or the math, it simply not in the same league. Although all thing NOT being equal, I thought I would bring it up only for a visual comparison to a higher bit count image. The M8, however, IMHO, does produce more engaging photos then the Nikon or the Canon.

[guywalder]

Remember as well that Nikon's compressed Raw stores more than twice as many values as the M8, but I haven't seen anyone say that the NEFs are better.

--HC

 

isnt that, at least to an extent, because Nikon are working in 8bit at the sensor and interpolating up to 16bit 'in camera', whereas Leica are working in 14bit at the sensor??

[AlanG]

The luma of the most saturated primary for instance (say RGB 255,0,0 for fully saturated red) is only one third the luma of white (RGB 255,255,255) since only one gun is firing. The same goes for the other primaries obviously. The fully saturated secondaries (say RGB 255,0,255 for magenta) are roughly twice as bright as the primaries give or take some variance in eye color sensitivity.

 

I'm not sure wher you are going with this or why it matters. When do we need the colors to be as bright as a white subject that is near by? Photography is an interpretation of the real world, not a re-creation of it. Thus we don't need a monitor that is capable of re-creating the light as it existed in reality. We only need to be able to make a reasonable represention of it.

 

Consider that in most photographic situations, we are recording light that reflects from the subject. Thus the visual response and photographic record will have similar results that will fit in with the monitor's design limitations. Most of the white light will be reflected from a bright white subject and most of the green and blue light will be absorbed by a red subject. Thus it is rare that subjects containing primary colors need to be depicted as brightly as the white that reflects from the subject. And in the case where this may be true - a very bright red neon sign next to a white sheet of paper for instance, one merely has to make the overall image darker and the red will look brighter in comparison. (The brain uses what is called "constancy" and adapts to the decrease in overall brightness.)

 

As for the 8 bit part. There aren't many monitors that display at a higher value yet.

[sdai]

That's interesting stuff, but what I wanted to see has nothing to do with resolution. It is the dynamic range and the ability to adjust the files. (Especially if the M8 is only working in 8 bit.)

If you're looking at DR then M8 is the champion among the three ...

 

At base ISO:

 

1Ds2 8.5 stops

DMR 9 stops

M8 9.5 stops

 

But this is not the point of the original question ... so people see M8 is already very good in this department, and I have no objection to that. My only suspicion is ... why can't it be even better?

 

The pictures I take may not be worth anything by many people's standards but as a gearhead, I want my camera beat everything both in practice and on paper ... and that's why I spend top dollars on Leica lenses even though they would make absolutely no difference to my pictures.

[sdai]

As of the RAW file torture test, there're at least two dozens of forum posters who claim they have both ... I don't see why they can't be bothered to do a quick test you've suggested.

[sdai]

I should have also mentioned that the 1Ds2 is not the best DR performer in Canon's lineup ... at base ISO, the 1D Mark II has 9.5 stops and the 5D has 9 stops. Now the 1D3 with 14bit processing I believe it can easily go beyond 10 stops.

 

Fuji's S3 already has 10 stops at ISO100 and 9.5 stops at ISO400 ... don't have the figures of the S5 yet, but I have little doubt that it can do better.

[ho_co]

isnt that, at least to an extent, because Nikon are working in 8bit at the sensor and interpolating up to 16bit 'in camera', whereas Leica are working in 14bit at the sensor??

Guy--

That may be.

 

I don't know really anything about how the Nikons work, but my guess would be that the D200, say, works from a greater than 8-bit depth.

 

It really is interesting to me how many 'solutions' have been attempted for image quality!

 

--HC

[sdai]

Nikon uses a 12-bit D/A in the D200, the compressed NEF stripped the color depth down to 10 bit ... and there's obvious difference naked eyes can easily detect already so I don't believe the myth that human eyes can't see it ... between 8 and 16 bit - that's one I put into my joke folder.

 

But then that's me ... when I used to frequent Fred Miranda's site, there were folks testing me with different combinations of Nikon cameras, telephotos and TCs. Not for once I have failed. LOL

[ennjott]

Amazing how this subject comes up again and again, along with lots of speculation presented as fact ("no camera can output true 16 bit color" - plain wrong).

 

The M8 look up table essentially causes a gamma encoding of the image. I modified dcraw so it will output the image without the LUT applied, and such a conversion looks indeed similar to jpegs straight from the camera.

 

Because most bits are used to represent the brightest tones, most of what you lose is in the "artificial film shoulder" that, when working in 12+ bits, allows us to get back details from what appears as perfect white when represented by 8 bits per color. What's probably also affected is the ability to completely change white balance, though I haven't tested that.

This is never visible without heavy image manipulation so it's not the end of the world like that reviewer seemed to think, but there's not the slightest doubt it's unworthy of a camera of that price. It's a crude fix for slow internal processing. It's also 1. not as intelligent as Nikon's compressed NEF (that yields higher effective bit depth and can be switched off in higher end models too) and 2. not comparable to Canon CRW/CR2 that uses lossless compression. Trying to explain that as some especially clever and new way to store images is simply ridiculous.

No one ever claimed the difference between 8 bits and higher color depth is directly visible (although there are worst case scenarios where this is indeed possible), the important property is headroom for editing. These are really only "half raws".

 

Writing images in the 12-bits JPEG mode that no other camera maker has yet used at least would have been something new...

[Advertisement]

Hallo,

Look here: Leica M

[guywalder]

the important property is headroom for editing. These are really only "half raws".

QUOTE]

 

Right on!

 

Simon,

I have no idea where you got the information that the D200 is using a 12bit A/D, Nikon only claim the same imaging engine as the D2x, and even Thom Hogan's 'techy' review, which talks about sensor design, makes no mention. I can't help thinking that Nikon would have been shouting about it if it really was the case (like Canon on the 1DIII for example)....

 

Guy