philipd : vision_spring06 : Bayesian Color Tracking
Bayesian Color Tracking

matlab code:

  • bayes_track( skinsamples, bgsamples, clip, cspace ):
    given vector of skinsamples, bgsamples, find the L0,L1,L2 layers for each frame of clip. Use cspace to transform from [ R G B ] to the appropriate colorspace. (defaults to identity)
  • [samps] = getRoi( img ):
    prompt for an roi input from img, returning a column of [ R G B ] triplets.
  • [P] = mvgaussian( p , mean, cov ):
    : using the multivariate gaussian formula listed, determine the probability for a column of values p, using mean and cov. mean and cov must have the same size as the entries in p
  • [colorout] = myhsv( colors ):
    rgb2hsv transform, normalizing from 255-scale doubles to 1.0 scaled doubles.
  • [colorout] = hsv2( colors ):
    rgb to hv2( in cylinder ): hue maps to theta, saturation to radius, and value is constant.

results