HMM-based Kalman snake for contour tracking

Hidden Markov model (HMM) provides a powerful probabilistic mechanism to incorporate multiple image cues, and can encode curve smoothness constraint in transition probabilities, therefore can be used to obtain more accurate measurement. Using HMM, the processing result is input into the Kalman snake filtering system as new measurement information, which can enhance anti-jamming capacity and tracking robustness of the filtering system. In the light of new inner product and norm definition of spline vectors, the normalization of shape matrix can furthermore improve the stability of filtering system and increase the system controllability of the model and parameters.