[sane-devel] Gamma tables causes distorsions : bug fixed ???

[Henning Meier-Geinitz]

Hi,

On Sat, Oct 19, 2002 at 12:05:40AM +0200, Aurelien Jarno wrote:
> A long time ago, there was a bug report submitted against sane 0.71-2 
> via the Debian BTS. It was a problem with custom gamma table which 
> causes undesired distortions :
> 
> | I have simply set the image intensity to about 1.5 and the red 
> | intensity to about 1.2.  Whenever some very bright part of the 
> | picture is scanned, it is not read in as white but rather as some 
> | other color such as a light blue, yellow, green, or red.  The white 
> | scanner background is almost a solid black.  This phenomenon persists 
> | both in the prevue and the final scan.

These kind of problems usually result from overflows in the gamma
code. E.g. for an 8-bit gamma table, 256 is written to a gamma table
entry unchecked. That's a bug that can happen either in the frontend
or backend.

> The answer was :
> 
> | This problem is known:
> | if you increase the size of the window in which the gamma-curve is  
> | displayed it could get better!

The answer was from Oliver Rauch, so maybe he can explain if it was/is
umax or xscanimage related.

> I have looked up in the changelog and in the source code, the mailing 
> list archives, and I haven't found any trace about a fix for this bug. 
> However, I failed to reproduce this bug on my computer.

If it is a xscanimage problem: There is still a TODO entry:

 - Gamma table seems to ignore maximum value. E.g. for an 8-bit (0-255)
   gamma table, the last entry may be set to 256 when setting the gamma
   value.

If I remeber correctly, this was an issue with the gtk gamma widget.
But it's a long time ago...

Bye,
  Henning