[Tig] jpg vs tiff, (OT), Glenn Chan

Bob Friesenhahn bfriesen at simple.dallas.tx.us
Wed Mar 12 15:34:08 GMT 2008


On Wed, 12 Mar 2008, Rob Lingelbach wrote:
>
> A NEF (Nikon RAW) image opened in The Gimp shows a certain amount of
> fine detail.  The same image opened in Cinepaint shows a lot more
> detail, in that a textured suitcase on the other side of the room in
> the shot has all of its texture visible in Cinepaint and none in The
> Gimp.   So I would guess, that bit depth in RGB (and I think the NEF
> file is 12 bits) contributes significantly to luminance detail.
> Forgive what might be a basic question, but I've not gotten into color
> bit-depth as applies to Mac/PC paint programs before.

I think that you are comparing apples to oranges.  Detail is primarily 
dependent on supported frequency response (i.e. filters & sampling 
frequency), which has little to do with bit depth.  If you are viewing 
the result on an 8-bit (or less) computer display, then that should be 
ample evidence that (to a certain point) bit depth does not have a 
large contribution to simultaneous displayed "detail".  It seems more 
likely that CinePaint has used better algorithms to convert from RAW 
to internal representation and then to display.  This does not 
surprise me since Kai-Uwe is fanatic about such details.  Since 
CinePaint imports a deep linear-light image, there is considerable 
opportunity to adjust the image within CinePaint.

Since GIMP only supports 8-bits internally, everything must be done 
while importing from RAW to 8-bit format (likely outside of GIMP using 
something like dcraw), with very little opportunity to improve things 
after the import is done.  With GIMP you would likely have to try 
several import iterations until you are happy with what has been 
imported.

Bit depth primarily helps with continuous tone images and color 
accuracy.

> Reading an article at http://photo.net/learn/raw/ , there is
> substantial reference given that for print work, 8 bits are enough.
> I have read elsewhere this is not true, rather that 8 bits are fine
> for Web, but 16 bit TIFFs are de rigeur for printing.    And to think
> of the math: 12 bit TIFF = 4096 intensity levels for each pixel.  8-
> bit (such as JPEG) = 256 intensity levels for each pixel.  And here is
> the quote from the the URL above that I'm curious about:

While I certainly agree that print and film require more depth, you 
are once again comparing apples to oranges.  The 12 bits in your NEF 
file use a linear-light representation while the 8-bits in the JPEG 
(and for computer display) use a gamma "corrected" value of 2.2. or 
2.4 so that each quantum step approximates a linear step in terms of 
intensity (what we see).  This means that each step in the 8-bit file 
is more efficient at storing the information.

The DCI XYZ format (for Digital Cinema) is also gamma encoded so that 
the bits encode the intensity most efficiently.

Bob
======================================
Bob Friesenhahn
bfriesen at simple.dallas.tx.us, http://www.simplesystems.org/users/bfriesen/
GraphicsMagick Maintainer,    http://www.GraphicsMagick.org/





More information about the Tig mailing list