[HDR-photo] glow around objects

[Geraldine Joffre]

On Sat, 24 Nov 2007 13:14:41 -0500, L.Yimm wrote
> Hi Geraldine,
> 
> Currently (at least for this project) my specific end result is a 
> tone mapped spherical image, taken from a spherical HDR created by 
> stitching 32-bit images.   And using 16-bit tiffs to make the HDRs 
> gives us a better HDR to stitch with.
> 
> When would it be advantageous to use the RAW images to create HDRs? 
I was thinking about the case when the source images were taken directly as
jpeg with a camera that applies sophisticated image enhancements when its
firmware processes sensors data. The cheaper the camera, the more likely there
will be additional processing to enhance contrast. The merge to HDR will then
have to find how to undo the contrast enhancements in order to recover the
linear values captured by the sensors. For that, it will have to guess what
the enhancement has been, which is not always trivial, and thus it may fail to
calculate HDR image values that are linear. 

Moreover, algorithms that attempt to guess what the firmware has done to the
linear data take usually as assumption that the operation was simply a curve
and that the same curve was applied to all exposures. Both assumptions are
unlikely to be valid with cheap camera models. So, with such cameras, it is
better to use the RAW files directly (assuming the camera offers a RAW mode)
to ensure the linearity of the HDR image data.

Geraldine Joffre