Jump to content

M8-why 10MB-vs-DMR 20MB


gogopix

Recommended Posts

Advertisement (gone after registration)

Exiftool output from a Canon cr2 file shows the following:

 

Bits Per Sample : 8 8 8

 

 

I'm not sure what that is, Jon ... but a quick run against a D2X file shows exactly that it has 12-bit color depth:

 

ExifTool Version Number : 6.48

File Name : DSC_0200.nef

File Type : NEF

Make : NIKON CORPORATION

Camera Model Name : NIKON D2X

Software : Ver.2.00

Subfile Type : Full-resolution Image

Image Width : 4288

Image Height : 2848

Bits Per Sample : 12

Compression : Nikon NEF Uncompressed

...

Link to post
Share on other sites

  • Replies 214
  • Created
  • Last Reply

Yes, I think there's more to the exif tag. From the exif spec at:

 

http://www.exif.org/Exif2-2.PDF

 

note this line:

 

Number of bits per component BitsPerSample 258 102 SHORT 3

 

and this line from a sample at:

 

EXIF Data

 

Bits Per Sample 0102 8 8 8 unsigned short 3

 

There it is again; 3 components for tag id 0102. The M8 and Nikon files reporting one number vs Canon's three. The definition in the pdf is:

 

BitsPerSample

 

The number of bits per image component. In this standard each component of the image is 8 bits, so the value for this tag is 8. See also SamplesPerPixel. In JPEG compressed data a JPEG marker is used instead of this tag.

 

Tag = 258 (102.H)

Type = SHORT

Count = 3

Default = 8 8 8

 

Oh well. All way beyond me!

Link to post
Share on other sites

Jon ... this is all the way beyond me too. :D

 

If my understanding is correct, BitsPerSample is a value paired with another one called Photometric Interpretation to decode the file. Canon obviously adopted a different class of values where their Photometric Interpretation setting refers to RGB, while the Photometric Interpretation value inside the Leica file and Nikon's NEF files are all set to Color Filter Array.

With the Photometric Interpretation set to CFA (Color Filter Array), the BitsPerSample value should be the color depth of the file.

 

Andrej, is this correct?

Link to post
Share on other sites

Guest guy_mancuso

Has anyone bothered to check the specs from the DMR sensor to the M8 sensor they are almost identical or even better the Exif data from a DMR to the M8 and look at that. There like kissing cousins .

Link to post
Share on other sites

Guy, you'd have never imagined that the best part is:

 

I HAVE A DMR!!! :D :D :D

 

ExifTool Version Number : 6.48

File Name : L2067892.dng

File Size : 19 MB

File Modification Date/Time : 2006:10:25 00:35:45

File Type : DNG

MIME Type : image/x-raw

X Resolution : 72

Y Resolution : 72

Resolution Unit : inches

Software : Rev 154(0.13.8.13.28.2.2)

Modify Date : 2006:10:24 11:49:20

Subfile Type : Full-resolution Image

Image Width : 3884

Image Height : 2592

Bits Per Sample : 16

Compression : Uncompressed

Photometric Interpretation : Color Filter Array

Strip Offsets : 194048

Samples Per Pixel : 1

Rows Per Strip : 2592

Strip Byte Counts : 20134656

Planar Configuration : Chunky

CFA Repeat Pattern Dim : 2 2

CFA Pattern 2 : 0 1 1 2

Default Crop Origin : 4 4

Default Crop Size : 3876 2584

Bayer Green Split :

Anti Alias Strength : 0

Version : 3.1

Raw File Name : L2067892.DNG

White Balance : As Shot

Temperature : 3950

Tint : +12

Auto Exposure : True

Exposure : -0.70

Auto Shadows : True

Shadows : 4

Auto Brightness : True

Brightness : 73

Auto Contrast : True

Luminance Smoothing : 0

Color Noise Reduction : 25

Chromatic Aberration R : 0

Chromatic Aberration B : 0

Vignette Amount : 0

Shadow Tint : 0

Red Hue : 0

Red Saturation : 0

Green Hue : 0

Green Saturation : 0

Blue Hue : 0

Blue Saturation : 0

Tone Curve Name : Medium Contrast

Tone Curve : 0, 0, 32, 22, 64, 56, 128, 128, 192, 196, 255, 255

Camera Profile : Embedded

Has Settings : True

Has Crop : False

Aperture Value : 3.4

Flash Fired : False

Flash Return : No return detection

Flash Mode : Unknown

Flash Function : True

Flash Red Eye Mode : False

Serial Number : xxxxxxx

Make : Leica Camera AG

Camera Model Name : R9 - Digital Back DMR

Date/Time Modified : 2006:10:24 11:49:20+02:00

Orientation : Horizontal (normal)

Creator Tool : Rev 154(0.13.8.13.28.2.2)

Rating : 0

Exposure Time : 1/12

F Number : 3.4

Exposure Program : Program AE

ISO : 100

Exif Version : 0220

Date/Time Original : 2006:10:24 11:49:20

Create Date : 2006:10:24 11:49:20

Shutter Speed Value : 0.9

Exposure Compensation : 0

Metering Mode : Multi-segment

Light Source : Other

Flash : No flash function

Focal Length : 0.0mm

Color Space : Uncalibrated

Contrast : Normal

Saturation : Normal

Sharpness : Normal

Image Unique ID : 0000000000000000004C764900000BA7

DNG Version : 1 0 0 0

Unique Camera Model : Leica DMR

Color Matrix 1 : 1.743042 -0.9219971 -0.05395508 -0.4549561 1.333008 0.1290283 -0.09301758 0.2259521 0.7819824

As Shot Neutral : 0.8112498 1 0.6125776

Camera Serial Number : xxxxxxx

Maker Note Safety : Safe

Raw Data Unique ID : 927BEB514CE8F3D6B8F03D61FA93C172

Aperture : 3.4

CFA Pattern : [Red,Green][Green,Blue]

Image Size : 3884x2592

Shutter Speed : 1/12

Focal Length : 0.0mm

LV : 7.1

Link to post
Share on other sites

Guest guy_mancuso

Advertisement (gone after registration)

Certainly have good taste.:D

 

 

So are the files in question the shipping firmware or the beta stuff, i got lost here was out all day actually shooting , nice change. LOL

Link to post
Share on other sites

Here's something interesting. Checking out these little snips from exiftool results...

...from a 20D .cr2

ExifTool Version Number : 6.49

File Name : _MG_2188.CR2

File Size : 6 MB

File Modification Date/Time : 2006:10:24 22:34:08

File Type : CR2

MIME Type : image/x-raw

Image Width : 1536

Image Height : 1024

Bits Per Sample : 8 8 8

Compression : JPEG (old-style)

 

... from the same 20D file converted to .dng

ExifTool Version Number : 6.49

File Name : _MG_2188.dng

File Size : 5 MB

File Modification Date/Time : 2006:10:24 23:34:38

File Type : DNG

MIME Type : image/x-raw

Strip Offsets : 37348

Rows Per Strip : 171

Strip Byte Counts : 131328

Artist : unknown

Image Width : 3596

Image Height : 2360

Bits Per Sample : 16

Compression : JPEG

 

...from a 5D .cr2

ExifTool Version Number : 6.49

File Name : _MG_9747.CR2

File Size : 11 MB

File Modification Date/Time : 2006:09:22 23:53:34

File Type : CR2

MIME Type : image/x-raw

Image Width : 2496

Image Height : 1664

Bits Per Sample : 8 8 8

Compression : JPEG (old-style)

 

... from the same 5D file converted to .dng

ExifTool Version Number : 6.49

File Name : _MG_9747.dng

File Size : 9 MB

File Modification Date/Time : 2006:10:24 23:34:46

File Type : DNG

MIME Type : image/x-raw

Strip Offsets : 109956

Rows Per Strip : 171

Strip Byte Counts : 131328

Image Width : 4476

Image Height : 2954

Bits Per Sample : 16

Compression : JPEG

Link to post
Share on other sites

So are the files in question the shipping firmware or the beta stuff, i got lost here was out all day actually shooting , nice change. LOL

 

Not sure if 1.06 will be the shipping firmware but it's definitely the latest known to people on earth. :D :D :D

Link to post
Share on other sites

Guest guy_mancuso

I don't think that number would be a shipping version to be honest , it is in nowhere land. They will most likely be like Version 1.0 than the first firmware update would be 1.1. Some kind of beginning . The DMR is version 1.1 than went to 1.2 and 1.3 is next. The 1.6 must be a beta version

Link to post
Share on other sites

Regrading tags will take a closer look: but I think cr2 is a cannon proprietary format, so they use only a part of exif tags, I found 2 bit depth tags in 350d file, both with 3 values=8.

 

As for DMR I have a support for it in my program so I know that the data is full 16 bit. The sensors a very different I suppose. Try to find DMR specs. You can see from my post with kodak specs M8 actually needs at least 12 bits.

 

It's seems that leica is worried, have you seen thats going on in 'DNG M8 files Online', do you connect it with our 'encoding' talks?

Link to post
Share on other sites

Ok some DMR vs. M8 SNR comparison:

 

M8 Kodak KAF 10500 SNR 71.5 db = 3758 levels = 12 bits

DMR Kodak KAF 10010 SNR 67 db = 2238 levels = 12 bits (but more close to 11)

 

According to specs M8 has a great chip saturation level is 1.5 x times more than DMR one, 60000 vs. 40000 (usual value for a 6.8 µm cells) electrons. Dont know how kodak guys managed to get such saturation level.

Link to post
Share on other sites

Guest guy_mancuso
Ok some DMR vs. M8 SNR comparison:

 

M8 Kodak KAF 10500 SNR 71.5 db = 3758 levels = 12 bits

DMR Kodak KAF 10010 SNR 67 db = 2238 levels = 12 bits (but more close to 11)

 

According to specs M8 has a great chip saturation level is 1.5 x times more than DMR one, 60000 vs. 40000 (usual value for a 6.8 µm cells) electrons. Dont know how kodak guys managed to get such saturation level.

 

 

This to me means more Dynamic range from the DMR , reason this is so baffiing to me. So somehow they figured a way to get more range from essential the same sensor. It is a different sensor but is very close to the DMR. that is why i said early there simply is no way to be 8 bit ,more DR and less bits just does not compute real well. Now hell if it is 8 bit and i gain 1.5 stops in DR , damn i may jump all over that. Overall just not make sense not saying your data is wrong or anything like that , there just is no logic to it. But i am far from a engineer .

Link to post
Share on other sites

This to me means more Dynamic range from the DMR , reason this is so baffiing to me. So somehow they figured a way to get more range from essential the same sensor. It is a different sensor but is very close to the DMR. that is why i said early there simply is no way to be 8 bit ,more DR and less bits just does not compute real well. Now hell if it is 8 bit and i gain 1.5 stops in DR , damn i may jump all over that. Overall just not make sense not saying your data is wrong or anything like that , there just is no logic to it. But i am far from a engineer .

 

Guy, I suppose you mean more DR from M8. And difference is not 1.5 stops, as you can see. Also dont forget what at first they get data as 14 bit.

Link to post
Share on other sites

Guy, I suppose you mean more DR from M8. And difference is not 1.5 stops, as you can see. Also dont forget what at first they get data as 14 bit.

 

Andrej, Assuming they are using a 14 bit DAC, that maybe just to avoid the 1/2 LSB error they would inevitably get with a 12 bit DAC and they may then be truncating the data to 12 bit after rounding to give a nicely dithered LSB and an accurate measure of the sensor output limited by the sensor's resolution, not the DACs.

 

If that is the case, mapping to 8 bit from 12 bit and back introduces much smaller errors than doing the same with 14 bit.

Link to post
Share on other sites

Mark, I dont know. We only know that 2^14 is the maximum value of back-transformation LUT. But here is some new info - such type of coding was used in imaging modules of several space missions. Some reasearchers stated that such mapping helps to match the quantization level to the photon noise (I've wrote about photon noise before) and helps to distribute quantization errors uniformly across the dynamic range.

Also there is some statements that such transform provided a dynamic range of 14 bits in the 8 bits.

 

Will look for more information, but here is the question, if it so simple why other companies dont use it, why they are bothering with a lossless compression, that you think?

Link to post
Share on other sites

But here is some new info - such type of coding was used in imaging modules of several space missions. Some reasearchers stated that such mapping helps to match the quantization level to the photon noise (I've wrote about photon noise before) and helps to distribute quantization errors uniformly across the dynamic range.

Also there is some statements that such transform provided a dynamic range of 14 bits in the 8 bits.

 

Andrej, if you can share the names of the researchers who have looked into the properties of this transformation for imaging data, I'd like to do some reading, too. (I am a physicist first, but make my living in computer science.) Noise in a CCD, in particular, comes in several flavors, some important in the dark, some important at higher temperatures, and some at higher light intensities, so modelling it and its impact on dynamic range is complex, but not impossible to understand.

scott

Link to post
Share on other sites

Archived

This topic is now archived and is closed to further replies.

  • Recently Browsing   0 members

    • No registered users viewing this page.
×
×
  • Create New...