FACE DETECTION (work in progress)

An attempt to detect faces in images. This is done by converting the image into YCbCr color space and then apply a kind of threshold to find pixels with skin colors. For more information how this is done, please see [1]. Implemented in C++.

original

Image 1: Original image.
segmented
Image 2: Original image segmented using the algorithm in [1] to find skin colors.
morphological operations
Image 3: Some morphological operations are applied with the purpose to remove irrelevant pixels.
skin areas
Image 4: Using the above image to mask out skin areas greater than 200 pixels.
skin areas
Image 5: Using image 3 to mask out skin areas greater than 200 pixels.
skin areas
Image 6: Using image 3 to mask out skin areas greater than 200 pixels.

Todo

  • lighting compensation
  • find facial features, determine if a face candidate really is a face


References
[1] Senior, A., Hsu, R., Mottaleb, M. A., and Jain, A. K. 2002. Face Detection in Color Images. IEEE Trans. Pattern Anal. Mach. Intell. 24, 5 (May. 2002), 696-706.
http://www.cse.msu.edu/~hsureinl/facloc/