Background Subtraction with Dirichlet process Gaussian Mixture Model (DP-GMM) for Motion Detection
Himani K Borse, Bharati A. Patil · International Journal of Scientific Engineering and Research · 2015
Video analysis often starts with background subtraction.This problem This problem is often loomed in two steps: Per-pixel background model followed by regulation scheme.A background model allows it to distinguished on Per-pixel basis from foreground, though the regularization combines information from adjacent pixels .Dirichlet process Gaussian mixture models is a method, which are used to approximate per-pixel background distributions followed by probabilistic regularization.Per pixel modes are automatically count by using non-parametric Bayesian method, avoiding over-/under-fitting.We implemented this method using FPGA and also compare the results with different methods like Background subtraction; Frame difference and Neural map and shows how this method is superior then previous methods.