Multiscale fusion of wavelet-domain hidden Markov tree through graph cut

Since object boundaries appear blurry, reduced localization accuracy of wavelet-domain hidden Markov tree-based (WHMT) method poses a problem during the object extraction process. A novel approach to improve localization accuracy by fusing multiscale information of the tree model is presented. We start with calculating the multiscale classification likelihoods of wavelet coefficients by expectation-maximization (EM) algorithm. Energy function is then generated by combining boundary term estimated by classification likelihoods with regional term obtained by approximation coefficients. Through energy minimization via graph cuts, objects are extracted accurately from the images. A performance measure for tobacco leaf inspection is used to evaluate our algorithm.