8 bits versus 16 bits

[jaapv]

Advertisement (gone after registration)

Then you would be looking at the effect of the algorithm used by photoshop to do 16->8bits with TIFF, here he is showing the impact of the algorithm used by Leica to go from 16->8 bits which is possibly different from the one used by PS to compress your TIFF (is PS method public ? Is it just stripping LSB ? Or using a SQRT as this ?). This would actually be a valid comparison to do, to see which ones creates the most artifacts.

 

All 16->8 bits conversions are not equal in term of visual impact, and all this is showing to me is that this method has a significant visual impact, but I'd like to see it compared to the "dumb" method of stripping LSB and eventually other existing converters.

That is correct, Leica applies a log curve to simulate a 12 bits file, despite the 8-bits compression. Rob's result till now, compared to Michael's similar experiment using Leica software makes me wonder if there is no additional manipulation which we do not know about

[diogenis]

Jaapv,

Since you do have both the DMR and m8, did you noticed that big of a difference as presented here? What's your opinion on that matter? Can you show us the same photo taken with DMR/M8?

edit: How about you try shooting scene with both cameras, which will cover both shades as well as highlights? like, say inside a room with bright windows showing outside surroundings

[Advertisement]

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

[jaapv]

As soon as I have time I will do just that. This time using the same lens, tripod and focus bracketing. In general I can say unscientifically that as good as the M8 is, the DMR seems to produce richer and more robust files.

[t024484]

Put it this way, Jaap, you can open up a 16 bit TIFF in Photoshop and immediately save it as a 8-bit TIFF, and DNG is based on TIFF so what are you proving from this?

 

With all respect, but I don't think that you do understand the process from sensor to picture.

 

1) Leica M8 digitizes the sensor output with a 14 bit convertor, then compresses each 14 bit value to 8 bits (according to a SQRT) directly after digitizing and writes it into the DNG file.

 

2) Leica DMR digitizes the sensor output to 16 bit, but does not compress the data, and writes is as 16 bits into the DNG file.

 

3) The DNG file holds RGRGR for one line and BGBGB for the next line.

 

4)When converting this Bayer scheme into a picture in your RAW converter, for each pixel the missing two colors have to be calculated from the surroundings by interpolation or averaging. If at this stage the data has been compressed, the interpolation will be hampered by the fact that all data is less accurate.

If the interpolation takes place with higher precision, the result will be much better.

 

5) When storing the data, after the interpolation to RGB has been done, you can choose to keep the data as 16 bit or as 8 bit not by SQRT compressing but by truncation or by rounding. Notice that each pixel now has three values, one for Red, one for Green and one for Blue. So reducing to 8 bits, in fact means reducing to 3*80 = 24 bits.

 

So storing a TIFF as 8 bit or as 16 bit has a complete different meaning as storing the sensor output as 8 or 16 bits,

And yes you can compress the 16 bit DMR data, like the M8 does, to get an impression whit this does to the picture quality.

[sandymc]

3) The DNG file holds RGRGR for one line and BGBGB for the next line.

 

No - the next line is GBGBGB. All Bayer sensors have the two green sensors on the diagonal.......

 

As a matter curiosity, how are the images being post-processed - raw converter, magnification, etc

 

Sandy

[delander †]

As soon as I have time I will do just that. This time using the same lens, tripod and focus bracketing. In general I can say unscientifically that as good as the M8 is, the DMR seems to produce richer and more robust files.

 

I will be very interested to see this comparison because I have often thought that M8 images are slightly grainy when compared to those from a Canon 5D. I have just thought that this is due to the lack of an AA filter or more noise sensor.

 

More grainy? Does that mean a less digital look more film like?

 

Blue skies from the M8 shot at 160iso show lots of graininess when displayed at 100%. Conversion to B&W, especially if you are applying a digital yellow or red filter, makes this grain even more prominent on screen. I have not studied how this affects the printed file.

 

 

Jeff

[jaapv]

Advertisement (gone after registration)

I will wait for optimal weather; blue sky to include and constant light.

Btw, I find that this "graininess" is more the screen than the file. I saw less of it on my previous NEC than on my current Eizo, and in print it somehow gets less as well.

[thrid]

I am more interested in maliability of the tone curve between 8bit and 16 bit raw file. I believe 8bit and 16bit makes more difference in this regard. I really do wish M8 had an option to record 16bit file.

 

Exactly.

[diogenis]

I will wait for optimal weather; blue sky to include and constant light.

Btw, I find that this "graininess" is more the screen than the file. I saw less of it on my previous NEC than on my current Eizo, and in print it somehow gets less as well.

 

thats another problem.... monitors...

[t024484]

No - the next line is GBGBGB. All Bayer sensors have the two green sensors on the diagonal.......

 

As a matter curiosity, how are the images being post-processed - raw converter, magnification, etc

 

Sandy

Have a look over here

RGB

[georg]

Interesting experiment and propably more valid than the comparison M8 vs. DMR before.

 

The LFI compared the M8 with 8bit-log to an M8-prototype with 16bit-lin output (that's the way Leica tested 8bit-log themselves and decided for it) and they printed a comparison (with bright areas, because this is the weakness of log, not the shadows) with no visible difference at all!?

Don't forget that besides the theoretical advantage of 16bit lin over 8bit log you still need a sensor/electronics that is able to fill these formats with data!

 

Somebody asked for blue skies @ 160ASA? Here you go: 160ASA -1/3EV, C1v4, no noise reduction, sharpened and converted to 16bit-TIFF - I hope it survived the final JPG-compression for web well.

 

@t024484

Leica claims that both, DMR and M8, are working with 16bit internal, while the M8 is using a 8bit-log-compression (which also has to be converted into a 16bit-TIFF).

How do you know that the DMR is using 16bit-converters and the M8 14bit-converters?

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[sandymc]

Have a look over here

RGB

 

 

Yes, like I said, the green sensors are on the diagonal, exactly as your reference shows them....

 

Sandy

[jaapv]

Interesting experiment and propably more valid than the comparison M8 vs. DMR before.

 

The LFI compared the M8 with 8bit-log to an M8-prototype with 16bit-lin output (that's the way Leica tested 8bit-log themselves and decided for it) and they printed a comparison (with bright areas, because this is the weakness of log, not the shadows) with no visible difference at all!?

Don't forget that besides the theoretical advantage of 16bit lin over 8bit log you still need a sensor/electronics that is able to fill these formats with data!

 

Somebody asked for blue skies @ 160ASA? Here you go: 160ASA -1/3EV, C1v4, no noise reduction, sharpened and converted to 16bit-TIFF - I hope it survived the final JPG-compression for web well.

 

@t024484

Leica claims that both, DMR and M8, are working with 16bit internal, while the M8 is using a 8bit-log-compression (which also has to be converted into a 16bit-TIFF).

How do you know that the DMR is using 16bit-converters and the M8 14bit-converters?

 

Yep- I see the same grain on my files - as I said I didn't on a midrange NEC monitor. But please present the file without sharpening - there are strong artefacts and halos.

[georg]

Yes, the jpg-compression is simply too strong, here's a extreme crop as a png, works propably better? I have sharpened it to enhance the noise a bit.

 

But this is normal "noise" I have found in every DSLR-file when using RAW and absolutely no noise reduction at all.

Welcome, dear visitor! As registered member you'd see an image here…

Simply register for free here – We are always happy to welcome new members!

[t024484]

Interesting experiment and propably more valid than the comparison M8 vs. DMR before.

 

@t024484

Leica claims that both, DMR and M8, are working with 16bit internal, while the M8 is using a 8bit-log-compression (which also has to be converted into a 16bit-TIFF).

How do you know that the DMR is using 16bit-converters and the M8 14bit-converters?

The M8 is using a SQRT conversion and not a LOG conversion.

The decompression table in the M8 DNG file, expands the data to 14 bit, that's why I think the A/D is also 14 bit, but this is no hard prove, just a logical interpretation.

Theoretically it could just as well be 12 bit or 16 bit.

But 16 bits would need a look up table that is 4 times as long, I do not know if the M8 has enough internal (flash)memory for this., and why starting with 16 bit, when you only expand it to 14?

A 12 bit convertor would be quite coarse.

 

I have both a complete black and a complete white DNG picture from the DMR supplied by Jaap..

White is represented as "FF FF", and black as "00 0X", where X is 0 most of the time. But because of sensor noise I can see here all figures ranging from 0 to F.

A 14 bit converter would only have shown 0, 4, 8 or C for the X value.

That's why I believe MJH's claim for 16 bits, He is usually very well informed.

 

Hans

[mjh]

The decompression table in the M8 DNG file, expands the data to 14 bit, that's why I think the A/D is also 14 bit, but this is no hard prove, just a logical interpretation.

Theoretically it could just as well be 12 bit or 16 bit.

But 16 bits would need a look up table that is 4 times as long, I do not know if the M8 has enough internal (flash)memory for this., and why starting with 16 bit, when you only expand it to 14?

A 12 bit convertor would be quite coarse.

Actually, 12 bit was considered state of the art until 2007 when DSLRs like the Fuji S5 Pro and Canon EOS-1D Mark III started to use 14 bits. The DMR was built by Imacon who were accustomed to resolving 16 bits in their MF backs; it was only natural for them to stick with the technology they were familiar with. Still, I would be surprised if the least significant two bits weren’t just noise.

 

The M8 may use a 16 or a 14 bit converter, but since Leica considers just 14 bits to be significant, it doesn’t really matter whether the A/D converter should resolve two additional bits, only to be discarded later. Effectively, it is a 14 bit A/D converter.

[ammitsboel]

Why isn't LSB dithering discussed?

In audio it has a severe impact to the audio that lands on a CD, if there was used dither when going from 24bit or higher down to 16bit and what kind was used. A truncated signal is often rough and flat.

 

I know that I must consider that everything doesn't relate to audio, but still I'm interested in how dither will affect the result.

[isaac]

Hello!

1) Leica M8 digitizes the sensor output with a 14 bit convertor, then compresses each 14 bit value to 8 bits (according to a SQRT) directly after digitizing and writes it into the DNG file.

Only in case the lensdetection (and UV/IR compensation) is OFF.

4)When converting this Bayer scheme into a picture in your RAW converter, for each pixel the missing two colors have to be calculated from the surroundings by interpolation or averaging.

And not to forget prediction.

[sdai]

Then you would be looking at the effect of the algorithm used by photoshop to do 16->8bits with TIFF, here he is showing the impact of the algorithm used by Leica to go from 16->8 bits which is possibly different from the one used by PS to compress your TIFF (is PS method public ? Is it just stripping LSB ? Or using a SQRT as this ?). This would actually be a valid comparison to do, to see which ones creates the most artifacts.

 

When the analog signals are converted into digital bits after the A/D, the DSP will do a series of pre-processing, such as demosaic, matrix, gamma correction, sharpening, etc ... before it reaches the point of constructing a RAW file whether compressed or not.

 

The DMR file are pre-processed and optimized with 16 bit output intended, the M8 files must be pre-processed with the 8 bit compression in mind to guarantee minimal artifact.

 

You don't know what type of pre-processing, what the amount of such pre-processing is applied to the DMR files and/or the M8 files before it's compressed or uncompressed. All this experiment does is to show what 16-bit to 8 bit compression with this particular method could do to a "cooked" DMR DNG file.

 

If you want to prove that 8 bit compression does harm to the quality of M8 DNG, you have to either get hold of the sensor dump, or at least a 16 bit M8 file before the compression. The M8 files may very well be pre-processed to offset the loss due to compression and minimize its artifacts. Who knows?

 

Have you seen a 16 bit M8 file? if not, then how can you prove the 8 bit file is worse than it?

 

Michael and Holger Sparr have played with a 16 bit prototype and both agree that the differences are trivial and insignificant (to say the least). And I believe what they say.

[Advertisement]

Hallo,

Look here: Leica M

[t024484]

When the analog signals are converted into digital bits after the A/D, the DSP will do a series of pre-processing, such as demosaic, matrix, gamma correction, sharpening, etc ... before it reaches the point of constructing a RAW file whether compressed or not.

 

I can only repeat that you seem to have no Idea what a camera is doing.

 

Hans