To reduce the difficulty of implementation and shorten the runtime of the traditional Kim-Fisher model, an entirely discrete Kim-Fisher-like model on lattices is proposed. The discrete model is directly built on the lattices, and the greedy algorithm is used in the implementation to continually decrease the energy function. First, regarding the gray values in images as discrete-valued random variables makes it possible to make a much simpler estimation of conditional entropy. Secondly, a uniform method within the level set framework for two-phase and multiphase segmentations without extension is presented. Finally, a more accurate approximation to the curve length on lattices with multi-labels is proposed. The experimental results show that, compared with the continuous Kim-Fisher model, the proposed model can obtain comparative results, while the implementation is much simpler and the runtime is dramatically reduced.
A parameter estimation algorithm of the continuous hidden Markov model isintroduced and the rigorous proof of its convergence is also included. The algorithm uses theViterbi algorithm instead of K-means clustering used in the segmental K-means algorithm to determineoptimal state and branch sequences. Based on the optimal sequence, parameters are estimated withmaximum-likelihood as objective functions. Comparisons with the traditional Baum-Welch and segmentalK-means algorithms on various aspects, such as optimal objectives and fundamentals, are made. Allthree algorithms are applied to face recognition. Results indicate that the proposed algorithm canreduce training time with comparable recognition rate and it is least sensitive to the training set.So its average performance exceeds the other two.