linear vs non linear tiff?

[nakleh]

Can anyone explain the difference between linear and non linear tiffs? I have been searching google for 2 hours and haven't come up with anything worthwhile..

I'm using Leaf 10 and C1PRO for processing, and searched both help files, and have come up with nothing also..

anyone?

[TheSteveMadden]

RAW files contain the original values from the individual pixels on a sensor. Our eyes percieve light in a non-linear logarithmic fashion, meaning that it takes 4 times as many photons for us to percieve a doubling of brightness. Linear conversion uses a Gamma of 1.0, instead of the normal 1.8 - 2.2 which looks natural.

Adobe has a pdf <<here>> (http://www.adobe.com/digitalimag/pdfs/linear_gamma.pdf) which explains Raw Capture and Linear pretty well.

[Edit] Corrected link above.

[Shasta]

Quoted directly form Fred Mirandas software site. Incredible stuff these guys make. Check out the stuff, I use his PS actions for my D30, unreal difference using the linear tiffs.

Why would you choose the linear 16-bit mode?

"RAW files converted to Linear 16-bit mode have broader dynamic range and color information (larger color gamut). In addition, linear 16-bit files are free of any sharpening algorithms and therefore images will have even lower noise levels then non-linear (normal) 16-bit images. The linear image provides image data without any processing with the full sensor dynamic range. The 16 bit’s file gives you 4096 points accuracy versus 255 in 8 bit images and therefore more room for post-processing preserving original dynamic range."

http://www.fredmiranda.com/software/

[nakleh]

Thanks for the response and explanation.. so I guess I'd HAVE to use Canon's processing software to create linear tiffs from my 5D's .cr2 files? I am currently using C1PRO and it seems it does not have this feature..

Jn.
http://www.nakleh.com

[TimothyFarrar]

The bigest advantage with processing in Linear vs a non-Linear colorspace is processing in Linear is color correct. When doing work on non-linear files, colors tend to come out slightly off. I believe that all Raw converters do all their processing in a Linear colorspace for this reason.

Your typical web colorspace sRGB and the common Adobe RGB 1998 colorspace are non-linear and have a gamma curve around 2.2.

Here is an example of what I am talking about when blending colors. See how the gray area between the green and magenta gets darker and darker the higher the gamma,

http://www.farrarfocus.com/ffdd/img/space0.jpg

However working in a Linear colorspace comes at a cost (expecially in Photoshop). For example, USM (unsharp mask) creates a really bad black outline around high contrast edges (there are workarounds for this however!).

Linear vs non-Linear (gamma) based colorspaces also effect the tonal distribution in the histogram. So when in Adobe RGB 1998 the middle gray color is close to the center of the histogram, this is not the case in Linear or say in ProPhotoRGB colorspace. Here is a cool graphic that shows an image in three colorspaces, the histograms, a red line for the middle gray shade, and the bands show the 9 zones of a simplifed Ansel Adams zone system,

http://www.farrarfocus.com/ffdd/img/space2.jpg


Anyone looking for the actual 8bit levels for the 9 zones and the middle gray shade can be found on this page towards the bottom,

http://www.farrarfocus.com/ffdd/zone.htm

Enjoy,