[HDR-photo] Why does the preview look better than the final output image?

[Geraldine Joffre]

On Tue, 27 Mar 2007 14:13:04 -0700, Bob Wise wrote
> Thanks Geraldine... I'm not quite sure I understand.
> 
> Are you suggesting to:
> 
> 1) Upres the image
> 2) Run the HDR processing
> 3) Down-res the image?

The suggestion of resizing the HDR image slightly was for the special case
when the final result is very dark and way different from the preview. In this
case, the suggested workaround would be as follows:
1) Generate the HDR image at the original resolution
2) Down-res the HDR image in Photoshop by a small amount
3) Tone map the resized HDR image

The purpose of resizing in this case is to get rid of the abnormally low
values (which are confusing the Details Enhancer algorithm) thanks to the
interpolation done when resizing.

For the general case of the final result looking different from the preview, I
was just pointing out that the algorithm is the same in both cases but the
image that is being tone mapped is different. The preview shows the result of
the tone mapping applied to a low-res version of the original HDR image. The
low-res version does not contain the same pixels as the original version of
the HDR image. Given that Details Enhancer takes into account local details
throughout the image, it is likely to map the low-res version of the HDR image
differently than its original version. 

The largest the difference in resolution, the more different the mapping is
likely to be. This is is why downsizing the HDR image to get closer to the
preview resolution should reduce the difference between the tone mapped
preview and the final tone mapped output. 

Another workaround is to use the tone mapping method Tone Compressor instead
of Details Enhancer. Tone Compressor maps each pixel independently of the
local context, making the preview a reliable representation of the final result.

Geraldine Joffre