4
$\begingroup$

I've been doing some background readings on demosaicing raw images. I've seen the Kodak PhotoCD raw image dataset mentioned in a few papers on demosaicing:

After some googling, I haven't been able to find the Kodak dataset. Where can I download the Kodak PhotoCD dataset?

Improve this question
$\endgroup$
2
  • $\begingroup$ Maybe this is what you are looking for? r0k.us/graphics/kodak $\endgroup$
    – Nicholas Kinar
    Commented Oct 24, 2012 at 0:05
  • 1
    $\begingroup$ If everything else fails, you can always ask the authors of the papers for the dataset they used. You have to be able to reproduce their results, and for that you need the dataset :) $\endgroup$
    – penelope
    Commented Oct 24, 2012 at 20:03

4 Answers 4

Reset to default
1
$\begingroup$

At this page you can find the link "PHOTO_CD" , and according to README file, folder "IMAGES" with "PCD" files contains RAW data of this images.

Improve this answer
$\endgroup$
3
$\begingroup$

The dataset appears to be in the public domain, and has been posted at this site here.

The pictures (24 in total) were originally released in 1991 on a CD. The original media with the actual image data could be requested via interlibrary loan; here is a WorldCat link to the repositories holding this item. The CD was originally distributed by Apple Computer. The raw version of the files do not appear to be available online.

Improve this answer
$\endgroup$
3
  • $\begingroup$ The PhotoCD images seem to be a benchmark for demosaicing raw images. Images that have not yet been demosaiced do not really have color...there's just one channel (instead of 3 channels) per pixel. However, the images on r0k.us/graphics/kodak have already been processed into multi-channel color images. Can anyone help me to find the completely raw version of the Kodak PhotoCD images? $\endgroup$
    – solvingPuzzles
    Commented Oct 24, 2012 at 0:46
  • $\begingroup$ @solvingPuzzles: I believe that the completely raw version of the Kodak PhotoCD images are on the actual disk. You might have to request it from another academic institution library. $\endgroup$
    – Nicholas Kinar
    Commented Oct 24, 2012 at 1:38
  • $\begingroup$ Sounds good. I just requested the PhotoCD on interlibrary loan. Once I get the data, I'll post it online (I doubt Kodak will come after me for this...). $\endgroup$
    – solvingPuzzles
    Commented Oct 24, 2012 at 1:54
2
$\begingroup$

http://r0k.us/graphics/kodak/ does link to Kodak PhotoCD contents with 2048x3072 images.

It is not what is called "raw" as those are using 4:2:0 subsampling (so not lossless at all) and xvYCC transfer function (almost like BT.709 gamma). Those were indeed produced by scanning KODACOLOR Gold 100 (35 mm) film and some other types of 35 mm, all types can be found here.

The problem is there are no correct decoder for those pictures: they are xvYCC, that means they are wide gamut, where superwhite and negative values specify extended color primaries stuff. While you can decode some part of it, to decode it all you need xvYCC TV and some decoder that does decoding of those values correctly. Worse variants of decoding are in ffmpeg and ImageMagic.

Native decoder from Kodak, of course, also does not support it, because back then there were no support of xvYCC in HDMI, that was added only in 1.3 part of the standard. Also on top of it icc color management can be done if the specified metadata requires that.

Improve this answer
$\endgroup$
1
$\begingroup$

RAW images do show color. Is green due to the double number of green cells of other basic components (RGB) having a Bayer filter or similar. It is on that where you have to apply the demosaicing algorithms to complete each color plane with values ​​in incomplete depending on their own neighbors pixels RAW image. nACHO (also looking those images)

Improve this answer
$\endgroup$
1
  • $\begingroup$ It seems that the Kodak PhotoCD images were produced by scanning 35mm film. Hence the RAW images in this case should contain all three color channels. While extremely high-resolution scan can reveal the underlying film "grain", this is not the case here - so there should be no mosaic pattern whatsoever in the RAW images. $\endgroup$
    – Boaz Arad
    Commented Oct 2, 2019 at 12:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.