7(i) RAW interpolation

[Patrick Porlan]

Are these 3 constants the same for every picture taken or do they
vary shot to shot?

These vary from shot to shot, and I don't know how their values are computed. Paul K. Hansen's MRW2TIF can print them. It's linked from:
http://webpages.charter.net/bbiggers/DCExperiments/html/minolta_raw_format.html

-- Patrick

 

[Peter Marina]

Hi Patrick and thanx for your prompt responce.

Concerning the interpolation algorithm, How do I know if I'm using variable gradiant or linear when pressing F3? ( the only sample mrw file I have shows very little difference even at 200% mag.)

Also you mention in the notes that you bundled with iShifter: "Each RGB component is coded using a 1+14+1 bits representation (for now). The two extreme bits are ignored by the display engine." I allways thaught that it was 12bit/color from the CCD. why do you mention 14bits?

Thanx ahead.

Peter Marina

Are these 3 constants the same for every picture taken or do they
vary shot to shot?

These vary from shot to shot, and I don't know how their values are
computed. Paul K. Hansen's MRW2TIF can print them. It's linked from:
http://webpages.charter.net/bbiggers/DCExperiments/html/minolta_raw_format.html

-- Patrick

--
Peter Marina

 

[Patrick Porlan]

Concerning the interpolation algorithm, How do I know if I'm using
variable gradiant or linear when pressing F3? ( the only sample mrw
file I have shows very little difference even at 200% mag.)

It's slower and the result is slightly sharper... There was a visual indication of the interpolation type but I removed it a few weeks ago. Keep in mind that there is no interpolation in reduced scales (i.e. anything but full size).

Here is the kind of difference that you can expect, with 800% magnification:





This is a little detail found in the extreme bottom-right of this D7 RAW picture:
http://perso.club-internet.fr/porlan/pict1013.zip

For those really interested, I'm sure that it would be interesting to see what DIVU 1, DIVU 2 and other MRW aware programs yield.

Also you mention in the notes that you bundled with iShifter: "Each
RGB component is coded using a 1+14+1 bits representation (for
now). The two extreme bits are ignored by the display engine." I
allways thaught that it was 12bit/color from the CCD. why do you
mention 14bits?

Right, the CCD data is only 12 bits. My note is about the internal representation used in my program. Internal computation is done using 16 bits numbers, and the mapping between the 12 bit sample value and the internal representation is int_val = sam_val * 8. I wanted reasonable precision without immediate clipping. Cryptic enough ?

-- Patrick

 

[Peter Marina]

Almost actualy... lets see if I still remember my computer basics....

you are adding 3 bits on the least significant side and adding a fourth one on the most significant side, all 4 new bits set to zero. But i still dont understand where "14" came from but hey.. I quit computer science for a reason...

Thanx for the info and the new raw file. If you are interested, I can post some comparisson pics between different raw converter, but I only have acces to the user-made software. Cheers

Peter Marin

Concerning the interpolation algorithm, How do I know if I'm using
variable gradiant or linear when pressing F3? ( the only sample mrw
file I have shows very little difference even at 200% mag.)

It's slower and the result is slightly sharper... There was a
visual indication of the interpolation type but I removed it a few
weeks ago. Keep in mind that there is no interpolation in reduced
scales (i.e. anything but full size).

Here is the kind of difference that you can expect, with 800%
magnification:





This is a little detail found in the extreme bottom-right of this
D7 RAW picture:
http://perso.club-internet.fr/porlan/pict1013.zip

For those really interested, I'm sure that it would be interesting
to see what DIVU 1, DIVU 2 and other MRW aware programs yield.

Also you mention in the notes that you bundled with iShifter: "Each
RGB component is coded using a 1+14+1 bits representation (for
now). The two extreme bits are ignored by the display engine." I
allways thaught that it was 12bit/color from the CCD. why do you
mention 14bits?

Right, the CCD data is only 12 bits. My note is about the internal
representation used in my program. Internal computation is done
using 16 bits numbers, and the mapping between the 12 bit sample
value and the internal representation is int_val = sam_val * 8. I
wanted reasonable precision without immediate clipping. Cryptic
enough ?

-- Patrick

--
Peter Marina

 

[AMonroy]

Thank you Patrick.

The D7i has not still arrived to my country (Spain), but when I
have it I could send you some file MRW.

I hope you warn me when you have a new version of Image Shifter.

A greeting,

AMonroy

Patrick, are you still interested in upgrading your program for the D7i? If you need files MRW I can send it. And if your you want that the pictures make them of some special configuration, tell it to me. The program is fantastic and I would be interested very much that was compatible 100% with my new D7i.

A greeting,

AMonroy

 

[Patrick Porlan]

Patrick, are you still interested in upgrading your program for the
D7i?

Yes

If you need files MRW I can send it.

Please try to post them on a file server. My mailbox is limited to a few MB. If you can't, drop me a mail and we'll see what we can do.

And if your you want that
the pictures make them of some special configuration, tell it to
me. The program is fantastic and I would be interested very much
that was compatible 100% with my new D7i.

No special configuration needed. Thanks,

-- Patrick