[Radiance-general] Average luminance calculation

Gregory J. Ward gregoryjward at gmail.com
Tue Jun 14 20:44:13 CEST 2005


Hi Jay,

You have to offer us some clues as to how you are attempting this  
calculation.  The website you give shows a bunch of JPEG thumbnails.   
Are you converting those?  It seems that the author of this page is  
converting them without considering gamma, then computing the log of  
the non-linear gray values, which is basically nonsense.  However, in  
the interest of reproducing their results, you can use:

     % djpeg egret2.jpg | ra_ppm -r -g 1 \
         | pvalue -h -H -d -b | rcalc -e '$1=log($1)' | total -m

Using the egret pictures, this then gives:

     -2.08354495

To get the final log-average luminance, simply apply the exponent  
function:

     % ev 'exp(-2.0835)'

This yeilds:

     0.124493721

Not exactly the 0.1212 value they give on the website, but close.   
Note that I didn't use a delta value in my calculation, but adding  
one should raise this value rather than lower it.  A more intelligent  
application of log averaging would apply a gamma to get back to a  
linear color space, e.g.:

     % djpeg egret2.jpg | ra_ppm -r -g 2.4 \
         | pvalue -h -H -d -b | rcalc -e '$1=log($1)' | total -m

Apply the exp() function to this result gives a much smaller result:

     0.00687406256

-Greg

> From: Jay <jpothara at uccs.edu>
> Date: June 14, 2005 10:37:17 AM PDT
>
> Hi people,
> I am trying to calculate contrast for images. I am referring to  
> paper (http://diuf.unifr.ch/courses03-04/imaging/simoncelli.pdf).  
> This paper suggest that to calculate contrast luminance is first  
> needs to be calculated.
> I am calculating lumninance using the formula.
> L_w = exp[ 1 / N( sum[ log( delta + L_w ( x, y ) ) ] ) ]
>
> Where:
>
>  * L_w - log-average luminance
>  * N - number of pixels in the image
>  * delta - a small factor to avoid problems with black pixels
>  * L_w ( x, y ) - the world space luminance for pixel ( x, y )
>
> where log has base e.
>
> I am using the image from http://www.cacs.louisiana.edu/~cice/lal/.  
> i cant get the luminance values mentioned on the website.
> Can someone please tell me where am i going wrong.
>
> thanks
> Jay



More information about the Radiance-general mailing list