Improving Face Tracking accuracy by previous image processing? RRS feed

  • Question

  • Hi,

    we would need improved accuracy on the head rotations (and animation units). The jumps between the values are often several degrees.

    My question is if the tracking output could be improved by doing image processing on the color or depth  pictures feeded into the facetracker? Like doing a histogram normalization on the color picture or reducing noise in the depth picture by a mean filter.

    Could you provide a little more information on which data is used to calculate the candide face model orientation. Is this solely based on the 100+ tracking points?



    Thursday, June 27, 2013 11:17 AM