A Full Causal Two Dimensional Hidden Markov Model for Image Segmentation

In this paper we propose a full causal two Dimensional Hidden Markov Model in which the state transition probability depends on all neighbouring states where causality is preserved. We have modified the Expectation Maximization algorithm (EM) for evaluating the proposed model. A novel 2D Viterbi algorithm is formulated to decode the proposed model with reduced complexity in decoding larger blocks. The proposed model can be used in areas such as image segmentation and classification. In particularly when applied to poor quality images such as ultrasound images with more ambiguous regions our model showed promising results when compared with existing models.