[HDR-photo] RAW to EXR

listmail at mab3d.com listmail at mab3d.com
Tue May 1 09:51:08 EDT 2007


On Apr 30, 2007, at 9:55 PM, Justin Johnson wrote:

> Is there a way to convert RAW files to HDR or EXR without applying  
> any adjustments during the conversion?  It seems that all the  
> aplications that do this raw conversion require that a white  
> balance be used.

<snip>
>
> Scene-referred by definition is the actual linear float values of  
> the scene.  If this is the case, and HDR images generated by these  
> applications are scene-referred, then how is it possible that they  
> allow for a white balance adjustment to be applied?
>
> It seems to me that applying any sort of color temp adjustment  
> would change the information and could no longer be called "scene- 
> referred".

Hi Justin,

It helps to remember that digital cameras (and their RAW formats) do  
not even pretend to be "linear" or scene-referred as capture devices.  
Cameras always assume there is some kind of lighting temperature, and  
will try to portray "pleasing and expected colors" within that light  
using everything from the lens coatings to the DA converters. The  
last part of this "color adjustment" is done when the RAW-processing  
software creates colors from the various filtered, grayscale sensor  
responses that the camera's electronics recorded. So, the decoding of  
a RAW file into a colored image requires *some* WB setting be applied  
during the conversion as a starting point for the creation of colors  
in the output file. (I think astrophotography has a method of using  
"no white balance" when converting RAWs shot without IR filters, but  
the files end up grayscale - no colors at all.)

Another thing is that scene-referred, in terms of HDRI, is best  
thought of as "trying to get closer to the actual values of the  
scene." The relative precision is based on an understanding of the  
capture device and its limitations, and combining multiple shots to  
"average out" things like dynamic range clipping (of course), noise,  
tight response curves, and even things like the color response of the  
bayer pattern of most sensors. HDR photography is still photography,  
and the capture device greatly influences the "accuracy" of what is  
captured  - especially when the capture device is designed to create  
"pleasing colors" above all else! If you instead use something other  
than a consumer camera (like a colorimetry spectrophotometer) then  
you can start with a more "objectively calibrated" assumption about  
the input values of your HDRs.

The "practical" bottom line: try to eliminate any unwanted effects of  
WB by keeping it consistent, and working within its limitations. You  
still have wiggle room there, but by combining multiple exposures  
into your HDRI file (as opposed to using a single input RAW) you can  
reduce relative "color errors" that the WB might impose when it  
shifts the RAW input to one end of the spectrum or another.

At least this is my understanding of WB and HDRI... I hope this  
helps, and I also hope that the more technically proficient on this  
list can chime in and correct me where I've gone wrong.

-Mark Banas

P.S. If Chad D. is still at Luma, tell him I say "howdy."



More information about the HDR-photo mailing list