[ptx] Adobe DNG format SDK

[Hal V. Engel]

On Tuesday 25 April 2006 09:44, JD Smith wrote:
> On Tue, 2006-04-25 at 12:46 +0100, Bruno Postle wrote:
> > On Tue 25-Apr-2006 at 08:26 +0200, Kai-Uwe Behrmann wrote:
> > > What are the problems, you meet with linear HDR's?
> >
> > I wasn't criticising cinepaint, I was really just repeating my
> > understanding from the thread JD Smith mentioned:
> >
> > http://thread.gmane.org/gmane.comp.graphics.panotools/38135/
> >
> > ie: Although it is possible to treat linear 16bit data as a kind of
> > 'limited dynamic range' HDR, this is a bit of a hack - 'Real' HDR
> > data is floating-point.
> >
> > Having said that, the 16bit approach works fine.  This page
> > describes it, but needs renaming to "16bit workflow with hugin":
> >
> > http://www.panotools.info/mediawiki/index.php?title=HDR_workflow_with_hug
> >in
>
> Part of my brainstormed workflow (apart from the MDR blending of 3
> +2,0,-2 bracketed images to fit snugly in a 16bit container) was the
> ability to work with the intermediate 16bit linear MDR images as if they
> were just normal images (TIFFs converted from RAW, say).  So if you need
> to mask out a moving person before running through enblend, just fire up
> your favorite 16bit editor and make it happen.  I think you couldn't do
> this with 32bit TIFFs, OpenEXR, or any of the other HDR formats.  I
> could never understand what was wrong with this proposed workflow, just
> some fairly vague warnings about linear editing being a false economy,
> which will do you more damage in the end than good.
>
> JD

I know that VIGRA works with 32 bit float TIFFs since I have tested this in my 
own software using images created in CinePaint.  CinePaint will work with or 
create TIFFs in 8 and 16 bit unsigned integer, 16 bit RnH short float, 16 bit 
fixed point 0-2.0 and 32 bit IEEE float per channel formats.   It appears 
that CinePaint will allow you to add an alpha channel to any of these formats 
but I have not tested this.   

VIGRA does not appear to support the 16 bit float or fixed point formats so I 
would not expect these to work with Hugin/enblend . I have not tested the 32 
bit float format with Hugin and/or enblend but I don't think it should be a 
problem.  In fact VIGRA supports images with up to 64 bit (double) float per 
color channel but I have not tested this in my application because I don't 
know how I would go about creating such an image.  But these might work in 
Hugin and enblend as well.  VIGRA also supports 32 bit integer (both signed 
and unsigned) per channel images so these should also work.

I would say that it is worth while to test Hugin and enblend with 32 bit TIFFs 
if you have software that will create these.

The above link to the panotools wiki has what appears to me to be a fairly 
conventional 16 bit work flow.  It is only slightly different from what I 
use.  But I do use UFRaw (so my images are gamma corrected) and I also color 
manage my images during raw conversion.  I also do not bother with down 
sampling to 8 bits/channel and when I do this would only be a copy of my 16 
bit image to be used for a specific purpose, such as display on a web page, 
where 8 bit/channel images are necessary. 

Hal