View Full Version : Motion Vectors and RSMB
01 January 2010, 08:51 PM
i got a problem with rsmb in after effects using motion vectors. blurs are appearing where they should not be. when i turn down the max displace value the effect becomes weaker but only disappears at zero. i used motion vectors within c4d 11.5 with a vector scale of 32. i am using 16-bit output.
here is a picture and an animation that shows the problem:
any idea? thanks
01 January 2010, 06:14 AM
Sorry, the info you provided is of no use. Please post sample frames from the RGB pass and the vector pass and tell us more about your render settings. It also seems to me that you are using DOF, which would be plain wrong when working with vector passes, as it messes up the evaluation of the motion samples.
01 January 2010, 09:32 AM
i uploaded a zip which contains the rgb and vector files and a .aep with the settings.
the dof was also a post effect in afx.
01 January 2010, 03:50 PM
Motion Vectors as PNG? That's never gonna work. The files must be rendered as 16bit or 32bit files (TIFF,PSD, EXR). This is also explained in the help files for RSMB. Therefore the artifacts you see must be considered "normal" or unavoidable. 8bit files simply don't have enough precision.
01 January 2010, 04:16 PM
i used png to make them smaller for upload, it is the same with openexr or tiff, i tried them all ^^
01 January 2010, 11:45 AM
I see. Then you will have to split up your scene into more passes and work with object IDs and only blur e.g. the raw diffuse pass. The problem I see with your setup is, that there is actually very little motion overall, resulting in these mismatches between the vector area and the RGB data. You can of course try to limit the influence by operating on the vector pass with tools like Minimax or other channel tools to contract the areas of influence. Don't get me wrong, such problems also exist when everything moves, but usually are less of a problem due to the perceptional psychology involved. For a fixed perspective like this, even rendering the elements separately and using the native 3D program motion blur seems a feasible option...
01 January 2010, 09:06 PM
thanks, i already thought about using object id's and will give it a try. i the final animation, there will also be some camerashake, maybe that will reduce the perception of the effect a bit.
01 January 2010, 12:13 AM
It might be the fact the motion vectors can't see through walls (it needs base pixels to smear across) unlike 3d motion vectors.
I believe the behavior of the plug is erratic when two objects moving in different directions on a same layer pass behind each other, inherent to the nature of the calculation
01 January 2010, 12:13 AM
This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.