For the current algorithms you have to pass your input images as grayscale. You can use cv::cvtColor to convert color images into their grayscale representation:
Algorithms like Eigenfaces and Fisherfaces require your training data to be of equal size. You can use cv::resize to resize your images:
Pay special attention to the documentation on cv::FaceRecognizer at:
Just look at the full source code listing on "Face Recognition in Videos" for example, because all preprocessing is done in there, too.
I am working on some algorithms for head pose estimation, but there's nothing I can release yet. Keep an eye on my blog,
Asked: 2012-09-11 14:43:41 -0600
Seen: 466 times
Last updated: Sep 11 '12
