Eight Bits?

[rsl]

Can someone explain why reducing the output of the M8 to 8 bits isn't a problem? Of course, if you shoot jpegs it's never a problem, but I've been shooting raw for years. Seems to me that reducing M8 output to 8 bits cripples post processing latitude.

 

[kbg32]

You might have your RAW processor adjusted wrong. I set the RAW processor in CS2 for 16 bit depth and color space to ProPhoto RGB. This gives me tremendous latitude. I also save a 16 bit tiff after conversion and adjustment. I adjust this 16 bit tiff to 8 bit only for printing or for submitting to an agency.

 

[Sailor Ted]

What is the difference between ProPhoto RGB and Adobe RGB? I always thought Adobe RGB was the way to go but wondered about ProPhoto RGB.

Thanks.

 

[rsl]

You might have your RAW processor adjusted wrong. I set the RAW processor in CS2 for 16 bit depth and color space to ProPhoto RGB. This gives me tremendous latitude. I also save a 16 bit tiff after conversion and adjustment. I adjust this 16 bit tiff to 8 bit only for printing or for submitting to an agency.

Keith,

Yes, you can convert any 8 bit file to 16 bits in Photoshop but you haven't gained anything. The 8 bit file has a potential 256 color variations per channel. When you convert to 16 bits you still have only 256 variations per channel even though the channel is capable of handling (not seeing or showing) 65,536 color variations per channel. You can't make a silk purse out of a sow's ear. Once the color depth has been reduced to 8 bits you can't recover what's lost. Even if you start out with a "16 bit" file you really don't have 16 bit depth because the camera doesn't deliver more than from 12 to 15 bits, but you have a lot more to work with.

 

[kbg32]

Ted,

ProPhoto RGB has a wider color gamut. Here is a link that should help explain.

http://www.luminous-landscape.com/tutorials/prophoto-rgb.shtml

Cheers,

Keith

 

[Sailor Ted]

Oh boy,
The Leica M8 has a non linear color bit depth. 16 bit in the shadows and mid tones, 8 bit in the highlights. This is to speed calculations in camera and they have provided some pretty complex arguments as to "why" over on LUF.

This is not an 8-bit camera- this is a non-linear 16-bit camera with 8-bit highlights. My first response was "huh, why?" But then I purchased one so I know (empirically) : )

 

[kbg32]

Russell,

I understand all that. To tell you the truth, I have not noticed the reduced color gamut at all. I think you might be more concerned then you have to be. The M8's color gamut is more than adequate for real world applications.

Keith

 

[rsl]

Russell,

I understand all that. To tell you the truth, I have not noticed the reduced color gamut at all. I think you might be more concerned then you have to be. The M8's color gamut is more than adequate for real world applications.

Keith

Keith,

I'm not trying to be contentious. I'm really asking because I want to know what Leica has done about color depth. As long as you follow Kodak's instructions for their early box camera: "Be sure there's plenty of sun, and be sure the sun's at your back when you shoot the picture," there'll never be a problem with file depth. The problem comes when you're shooting against backlight, for instance, without fill flash, and need to bring the shadows up in Photoshop without creating a noisy mess.

What Ted just said may be an answer, but earlier this morning I went to B&H's web site to see if they have M8s in stock yet (they have as of this writing), and read a review by a guy who did some shooting with the camera. This is what he had to say:


"The prototype for this camera had 14 bit output. The shipped version has
a vodo/compressed DNG of a very low 8 bits of color depth. Supposedly
the 14 bits the senor can capture is compressed via some special
compression and those 6 bit which are thrown away don't matter. This is
contrary to everything I have been taught about the value and
importance of capture images in the highest bit depth possible that and
processing this hit bit raw in 16 bit mode to ensure the highest output
possible. My teachers have been Jeff Schewe just to name of the seminal
figures and founding fathers of the digital.

"Additionally Leica has even admitted, albeit in long winded, in their
January and February issues of LFI magazine that their decisions to use
8 bit files does adversely affect image quality. Read the conclusion in the
LFI article for February and it is right there."


You can do all the compression you want to, but 8 bits is still 8 bits and 14 bits is still 14 bits. I don't have access to LFI magazine so I don't know what was in Leica's long-winded explanation and therefore can't judge whether or not it makes sense. I'm hoping someone can enlighten me.

 

[ClayH]

Leica are using a compression algorithm to compress 14 bits of data into an 8-bit number. My understanding is that this is through an exponent type of compression approach. In any case, what you will end up with is (assuming the best algorithms) a sample that retains the 14 bit range, but only has 8 bits of precision. There is no free lunch. FWIW, in practice, the files are fine and seem to be comparable to those I get from a 5D.

 

[jaapv]

It might be a good idea to start reading LFI. It has covered this topic exhaustively - and I even understood some parts of it.

 

[willie_901]

As mentioned above, ProPhoto RGB is a larger color space than Adobe RGB.

However, ProPhoto RGB looks flat on web sites. Most web sites are compatible with the sRGB color space. This fact caused a great deal of fear and loathing during the Adobe Lightroom beta test period.

In my workflow I use 4800 ppi, 16 bit tif images in ProPhoto RGB. When I'm satisfied with my Lightroom adjustments, I export a high-quality 8 bit sRGB jpg for web postings. For prints, I export to the color space and format my lab prefers (8 bit jpegs in Adobe RGB).


I would comment on the 8 bit M8 images, but I am fearful of immediate expulsion with no possibility of reinstatement, so I'll pass.

 

[icebear]

LFI explanation of 8 bits

LFI explanation of 8 bits

I bought the LFI maganzine yesterday and just briefly flipped through it. The article about the reduction to 8bits is quite lenghty and tries to explain the reasoning for Leica to go this route - mainly because of speed of writing the data to the storage card and saving storage space.

It's quite complicated and I just try from my memory as I don't have it handy right now. The key is that the data avaliable per pixel above and below the optimum exposure value are rated more important in the shadows than in the highlights. Supposedly giving you more detail definition where it counts. They bring up some examples with 2, 4, 8, etc. candles to explain the logarithmic scale for f-stop / time / ev and a waterhose with wholes in it to visualize the overflow/waste of data that doesn't give a real advantage. You can somewhat follow the idea behind the concept but there is no proof of principle shown. Meaning there is no direct comparison between a true 14bit M8 version and the actual 8 bit ("non-linear" 14bit).

 

[kbg32]

Russell, as I said, I think you are worrying needlessly. The files that this camera produces at the low ISOs are comparable to my 5D - smooth and buttery. I have seen no problems in the inherent bit depth from the M8's chip. The chip seems to act as well as the one in the 5D. I think a lot of people get too involved in what is written on paper. The M8 is a wonderful tool and like any tool has its pros and cons.

 

[rsl]

I export a high-quality 8 bit sRGB jpg for web postings. For prints, I export to the color space and format my lab prefers (8 bit jpegs in Adobe RGB).


I would comment on the 8 bit M8 images, but I am fearful of immediate expulsion with no possibility of reinstatement, so I'll pass.

Creating 8 bit files for printing makes sense. If you're printing to an inkjet the driver is going to convert the file to 8 bit anyway.

But with respect to your second comment, are you talking about 8 bit .JPEG or 8 bit .DNG? The statement has an ominous sound.

 

[rsl]

Russell, as I said, I think you are worrying needlessly. The files that this camera produces at the low ISOs are comparable to my 5D - smooth and buttery. I have seen no problems in the inherent bit depth from the M8's chip. The chip seems to act as well as the one in the 5D. I think a lot of people get too involved in what is written on paper. The M8 is a wonderful tool and like any tool has its pros and cons.

Keith, I'm used to D2X files. You may well be right, but before plopping five grand down for an M8 I want to know the straight skinny on things like bit depth and compression.

What Ted said earlier makes sense in a way. Since the sensor's response is linear, the vast majority of data is in the high end of the brightness range. According to Bruce Fraser in his "Real World Camera Raw" books, if you have a latitude of seven stops, half of the captured data is in the highest stop, half of the remaining half is in the next stop, etc., etc., to the end. That's why it isn't wise to get so worried about blowing out highlights that you underexpose excessively. To me that means that it would be great to have a way to add data to the lower part of the capture. Unfortunately, that's not possible. The next best thing is to avoid losing anything from the lower part of the capture. That might be possible. But the best thing of all is not to lose any data from any part of the capture. It looks to me as if Leica has opted for the next best thing, though I'd like to have some idea how they did that. I might add that trading quality for storage speed doesn't really light my fire. If I'm going to shoot bursts, I'm going to use my D2X that'll give me five 12.4 megapixel raw frames a second. With the Leica I'm going to do street work where you get one shot and that's it.

I'm glad to hear that you're happy with the images you get from the M8. Ted raves about them and the stuff of his I've seen backs him up. Only problem is that what I'm looking at is sRGB jpegs and that doesn't tell me what I really want to know. I did download a fullscale jpeg file of his and it was very good. But it's still jpeg and I'm concerned about what a backlit .DNG looks like before post processing.

 

[jaapv]

I did download a fullscale jpeg file of his and it was very good. But it's still jpeg and I'm concerned about what a backlit .DNG looks like before post processing.

If you pm me your e-mail I can put such a file on yousendit (at least try to; it would be the first time for me to use that service)

 

[rsl]

If you pm me your e-mail I can put such a file on yousendit (at least try to; it would be the first time for me to use that service)

Jaap, You can email me through the messaging feature on this forum or you can email me at rsl@rslstudio.com or rsl@pkinfo.com. Give me a URL for a zipped, .DNG from the M8 and I'll download it. I don' know what the size of a fullscale .DNG from the M8 is. It's about 19 meg for a .NEF from my D2X and about half that for one that's been converted to compressed .DNG, so, since the M8 has only 10 megapixels it's probably a bit less than that. I appreciate the offer.

 

[DaveSee]

Keith,

I'm not trying to be contentious. I'm really asking because I want to know what Leica has done about color depth. As long as you follow Kodak's instructions for their early box camera: "Be sure there's plenty of sun, and be sure the sun's at your back when you shoot the picture," there'll never be a problem with file depth. The problem comes when you're shooting against backlight, for instance, without fill flash, and need to bring the shadows up in Photoshop without creating a noisy mess.

The Leica DNG makes use of data as EXIF tag "LinearizationTable" which maps the 14bit data through this "filter" to yield the 8bit file data: together, the data + calculations using this matrix reconstitutes the "image" data in the RAW converter. A "16bit" file is actually 14bits, with 2bits of similar, but less computationally intense metadata as the M8 DNG... when reading the file from storage.

Rather than taking the time(and power) to record to SD card a 16bit file, Leica/Jenoptik chose to pre-process the data to create two "files"(faster, and less battery/power): one the data, and one the file metadata describing how each of 256 ranges of bits should be interpreted. Why faster and less power... why "sacrifice" data read /on/ the sensor? The combined metadata and 8bit data is smaller than a 16bit file, so the expensive disk(SD card) write is reduced. As for the algo to generate the map file: yes, it will lose some (truely)raw data, but no information that a 16bit capable image utility could use, reliably.

What Ted just said may be an answer, but earlier this morning I went to B&H's web site to see if they have M8s in stock yet (they have as of this writing), and read a review by a guy who did some shooting with the camera. This is what he had to say:


"The prototype for this camera had 14 bit output. The shipped version has
a vodo/compressed DNG of a very low 8 bits of color depth. Supposedly
the 14 bits the senor can capture is compressed via some special
compression and those 6 bit which are thrown away don't matter. This is
contrary to everything I have been taught about the value and
importance of capture images in the highest bit depth possible that and
processing this hit bit raw in 16 bit mode to ensure the highest output
possible. My teachers have been Jeff Schewe just to name of the seminal
figures and founding fathers of the digital.

"Additionally Leica has even admitted, albeit in long winded, in their
January and February issues of LFI magazine that their decisions to use
8 bit files does adversely affect image quality. Read the conclusion in the
LFI article for February and it is right there."

You can do all the compression you want to, but 8 bits is still 8 bits and 14 bits is still 14 bits. I don't have access to LFI magazine so I don't know what was in Leica's long-winded explanation and therefore can't judge whether or not it makes sense. I'm hoping someone can enlighten me.

The M8 writes an 8bit JPEG(with your choice of 3 colorspaces: sRGB, Adobe1998, or ECI RGB), and a raw, non-colorspace embedded/adjusted, DNG file with an 8bit payload and 8bit "converter" to yield, effectively the 14bits, however "not really RAW, dude".

You don't need LFI... just grab the DNG from Leica's website and see for yourself.

hth+rgds,
Dave

 

[Artichoke]

IMHO Leica did a fine job with their DNG engineering

IMHO Leica did a fine job with their DNG engineering

by using a complex system to compress output from the ADC, they saved a great deal of space with minimal sacrifice in file depth
I have a Fujifilm S3 that uses a 14 bit ADC and has the widest DR of an presently available digital camera, but it writes a 25 MB RAW file
fortunately Fujifilm did a superb job with its jpg compression engine, but the size of the files essentially cripple the S3 as a RAW camera
clearly with well engineered compression you lose little in image quality & I think Leica made an excellent decision with their DNG engineering
these files are very robust with a great deal of room in the shadows ...unlike the S3, which I learned to shoot for shadows as it holds highlights so well, even shooting jpgs, I expose the M8 for highlights and have been pleased with how well I can recover clean shadow data
the lesson I have seen with my S3 experience & now from the M8, is how critical the quality of the ADC data going into the compression is for the output of the file
garbage in means garbage out & if you present well engineered compression software with superior data (14 bit ADC for the S3 and I believe 16 bit for the M8,) you will get a superior file to cameras converting at 12 bits
of course the details are not known to me, but the results are clear ...I would not worry about the depth of the DNGs produced by the M8 ...these files will stand up to a beating quite well
what matters is what goes into the DNG file ...Leica (and Kodak/Jenoptik) packaged it sensibly to maximize depth and economize on space

 

[jaapv]

Jaap, You can email me through the messaging feature on this forum or you can email me at rsl@rslstudio.com or rsl@pkinfo.com. Give me a URL for a zipped, .DNG from the M8 and I'll download it. I don' know what the size of a fullscale .DNG from the M8 is. It's about 19 meg for a .NEF from my D2X and about half that for one that's been converted to compressed .DNG, so, since the M8 has only 10 megapixels it's probably a bit less than that. I appreciate the offer.

Russel. I'll work on it this weekend.