DNG Recover Edges with DMR

[ho_co]

Advertisement (gone after registration)

Since the M8 has an announced crop factor that is less than that of the DMR, it should follow that the M8 sensor is larger than that of the DMR. However, looking at Kodak's sensor offerings, nothing leaps out as having a diagonal of 32.5mm (M8) as compared to 31.7mm (DMR).

 

Another possibility is that the M8 might use the same chip as the DMR, but include edge pixels that the DMR doesn't access.

 

Therefore this question: When you use DNG Recover Edges (available at http://www.luminous-landscape.com/contents/DNG-Recover-Edges.shtml) with the DMR, what final image size (H pixels x V pixels) do you end up with?

 

Thanks for indulging my speculation.

 

--HC

[pascal_meheut]

I just tried. No difference.

[ho_co]

I just tried. No difference.

Hmm.

Should uncover several thousand new pixels.

 

You're so quick to respond, Pascal, let me ask the obvious: When you dragged the DNG file onto the applet's icon, did you get no message saying "program uncovered 72,332 pixels" or something like that? The difference to the eye will be minimal since you're only adding a couple rows of pixels on each side of the image.

 

Thanks for trying!! There goes that theory...

 

--HC

[pascal_meheut]

No message at all. But I can try again tonight, I did not check up to the pixel level.

[ho_co]

No message at all. But I can try again tonight, I did not check up to the pixel level.

No message probably means the program didn't run.

 

When I run DNG Recover Edges on a file on which it has already been run, I still get a message, in this case "Unable to recover any edge pixels from fileXXX.dng."

 

Thanks for your efforts, Pascal!

 

--HC

[pascal_meheut]

You are right. I tried again and it says "recovered 51.7 thousand pixels...". No big difference in the files as expected: 3884 pixels wide instead of 3876

[ho_co]

Advertisement (gone after registration)

You are right. I tried again and it says "recovered 51.7 thousand pixels...". No big difference in the files as expected: 3884 pixels wide instead of 3876

Thanks again for checking, Pascal!

 

That's really rather interesting, a much smaller recovery than on the D200, which yields 166.2 thousand pixels, going from 3872 to 3900 pixels wide.

 

So the DMR actually uses a higher percentage of available pixels on the chip than at least the Nikon D200.

 

Once again, you get what you pay for with Leica!

 

--HC

[ho_co]

If you're a Canon user, I'd be curious about your results with DNG Recover Edges as well.

 

I'm surprised about the efficiency of the DMR's usage of available pixels and would like more comparison data before developing my next theory .

 

--HC

[pascal_meheut]

I tried with a raw file from a Canon 350D (Rebel Xt) converted to DNG. Recover 76,000 pixels going from 3456 pixels wide to 3474.

[ho_co]

I tried with a raw file from a Canon 350D (Rebel Xt) converted to DNG. Recover 76,000 pixels going from 3456 pixels wide to 3474.

Very interesting, though I don't know whether it means anything.

 

From these examples:

 

The DMR uses all but 4 rows of pixels on each edge of the frame.

 

The 350D/Rebel XT uses all but 9 rows of pixels on each edge of the frame.

 

The D200 uses all but 14 rows of pixels on each edge of the frame.

 

So the Canon uses five fewer rows than the DMR, and the Nikon five fewer rows than the Canon.

 

Thanks for the information! (If I figure out what it means, I'll post it here first! )

 

--HC

[krabat]

Interesting discussion.

 

What about the quality of the additional pixel rows?

 

Regards, Peter.

 

P.S.: Indeed, when I "develop" my RAWs taken with the Digilux 2 with the RawShooter, I obtain pictures of 2566 x 1926 pixels. If I shoot in the jpg mode, the images are 2560 x 1920, i.e. six pixels smaller...

[pascal_meheut]

I saw no difference in quality. But of course, I did not do any formal test.

[ho_co]

Interesting discussion.

 

What about the quality of the additional pixel rows?

 

Regards, Peter.

 

P.S.: Indeed, when I "develop" my RAWs taken with the Digilux 2 with the RawShooter, I obtain pictures of 2566 x 1926 pixels. If I shoot in the jpg mode, the images are 2560 x 1920, i.e. six pixels smaller...

Peter--

There's a lot of things I don't understand about these issues.

 

The page where DNG Recover Edges is available mentions some of the factors, but the DMR has the smallest 'unused edge-pixel' count I've seen, and the spread seems to run just as I'd expect if you asked me to list brands in order of image quality.

 

In regard to the edge pixels, as Pascal said, there is no noticeable quality fall off. You're dealing with such a minuscule difference in image size that I think you'd have to have very sophisticated tools to measure a difference.

 

BTW: Using Adobe Camera Raw on the Digilux 2 files gives 2552 x 1920 pixels for RAW, as compared to 2560 x 1920 for JPG. DNG Recover Edges expands the image to 2568 x 1928 pixels, or four more rows on each short side, eight more on each long side. Seems that your RawShooter has already done most of that work!

 

So now it looks as if we've got another variable: the software used for the conversion.

 

Sigh.

 

--HC