• Create Account

Need scary sound effects or creepy audio loops for your next horror-themed game? Check out Highscore Vol.3 - The Horror Edition in our marketplace. 50 sounds and 10 loops for only \$9.99!

### #ActualGeometrian

Posted 30 July 2012 - 03:04 PM

Awesome, thanks!

I've implemented the bidirectional path tracer as described. It works in the sense that an image develops. However, the results are visually inaccurate:

Based on some thought and what I'm seeing, I think the problem is in the way the algorithm weights samples. Because the camera and light subpaths are importance sampled, they are not attenuated each bounce. Because there are far more paths of longer length than of shorter length, and the longer paths are not attenuated, the longer paths end up contributing more to the image.

EDIT: Actually, they should be attenuated by 1/pi for each bounce for diffuse surfaces. Hmmmm, that might do it.
EDIT-EDIT: After adding in an attenuation of 1/pi at each diffuse bounce, the image improves, but still looks similar.

Ideas? Thanks,
Ian

### #2Geometrian

Posted 30 July 2012 - 03:00 PM

Awesome, thanks!

I've implemented the bidirectional path tracer as described. It works in the sense that an image develops. However, the results are visually inaccurate:

Based on some thought and what I'm seeing, I think the problem is in the way the algorithm weights samples. Because the camera and light subpaths are importance sampled, they are not attenuated each bounce. Because there are far more paths of longer length than of shorter length, and the longer paths are not attenuated, the longer paths end up contributing more to the image.

EDIT: Actually, they should be attenuated by 1/pi for each bounce for diffuse surfaces. Hmmmm, that might do it.

Ideas? Thanks,
Ian

### #1Geometrian

Posted 30 July 2012 - 02:58 PM

Awesome, thanks!

I've implemented the bidirectional path tracer as described. It works in the sense that an image develops. However, the results are visually inaccurate:

Based on some thought and what I'm seeing, I think the problem is in the way the algorithm weights samples. Because the camera and light subpaths are importance sampled, they are not attenuated each bounce. Because there are far more paths of longer length than of shorter length, and the longer paths are not attenuated, the longer paths end up contributing more to the image.

Ideas? Thanks,
Ian

PARTNERS