Re: 16 bit per sample support

Jonathan A. Buzzard (jab@hex.prestel.co.uk)
Mon, 05 Apr 1999 23:06:33 +0000

njl98r@ecs.soton.ac.uk said:
> If you really must have back the 12-bit data you can do the second
> transform I showed, and get back the 12-bit data. There is no data
> loss involved. For most applications it is appropriate to use the full
> dynamic range, but you are right that some software will lose data
> working on 16-bit images. IMHO this software is broken (but I would
> say that because the tools I'm using don't have this problem)

Converting a 300MB image back to a packed 12bit image is a real pain,
especially if you have several. I don't agree that software that loses
data is broken, it is generally making a compromise to save on memory.
Frankly promoting all my images to 32 bit when I don't need to is
not an attractive proposition.

> If this level of accuracy (probably beyond human vision) is required
> for your application, you should suspect a lot of other components
> (optical problems at capture stage, development process, scanner
> hardware, ADC) before SANE.

Who said anything about human vision, I distinctly talked about processing
of images for quantitative results. If you think 12bit data is beyond
human vision you are simply wrong.

> The feature you describe (an additional toggle setting for stretch vs
> truncate) seems excessive even for Advanced options, where it will
> probably confuse users in a less technical environment. I don't think
> we want the advanced setting to mean "Deep magic for image processing
> people".

Throw hands up in air, so SANE is not for use by image processing people
then.

JAB.

-- 
Jonathan A. Buzzard                 Email: jab@hex.prestel.co.uk
Northumberland, United Kingdom.       Tel: +44(0)1661-832195

--
Source code, list archive, and docs: http://www.mostang.com/sane/
To unsubscribe: echo unsubscribe sane-devel | mail majordomo@mostang.com