Improved behaviour of media with respect to large distances or high densities (method 3 only).

Author: c-lipka

changes.txt

@@ -373,6 +373,8 @@ Reported via Windows Crash Reports:
 
 Miscellaneous:
 
+  - Fixed media cutting out or fading to black above the horizon (sample method
+    3 only; other methods remain problematic).
   - Animation status messages now consistently display a running frame count and
     the total number of frames actually to be rendered, properly taking into
     account the Frame_Step setting.

source/core/material/media.cpp

@@ -457,7 +457,18 @@ void MediaFunction::ComputeMediaAdaptiveSampling(MediaVector& medias, LightSourc
             // keep a sum of the results
             // do some attenuation, too, since we are doing samples in order
             // TODO - we could do even better if we handled attenuation on a per-sample basis
-            i->te += Result * Exp(-(i->od + ODResult * 0.5) * dd);
+
+            // Compute attenuation due to earlier sub-intervals.
+            Result *= Exp(-(i->od) * dd);
+            // Compute attenuation due to the sub-interval itself.
+            // NB: This formula is mathematically precise under the presumption that the ratio of emission to absorbtion
+            // remains constant throughout the entire sub-interval.
+            for (int iChannel = 0; iChannel < Result.channels; ++iChannel)
+            {
+                if (ODResult[iChannel] != 0.0)
+                    Result[iChannel] *= (1.0 - exp(-ODResult[iChannel] * dd)) / (ODResult[iChannel] * dd);
+            }
+            i->te += Result;
 
             // move c1 to c0 to go on to next sample/interval
             C0 = C1;