A truly 2-D hidden Markov model for off-line handwritten character recognition

In recent years, there have been several attempts to extend one-dimensional hidden Markov model (HMM) to two-dimension. Unfortunately, the previous efforts have not yet achieved a truly two-dimensional (2-D) HMM because of both the difficulty in establishing a suitable 2-D model and its computational complexity. This paper presents a new framework for the recognition of handwritten characters using a truly 2-D model: hidden Markov mesh random field (HMMRF). The HMMRF model is an extension of a 1-D HMM to 2-D that can provide a better description of the 2-D nature of characters. The application of HMMRF model to character recognition necessitates two phases: the training phase and the decoding phase. Our optimization criterion for training and decoding is based on the maximum, marginal a posteriori probability. We also develop a new formulation of parameter estimation for character recognition. Computational concerns in 2-D, however, necessitate certain simplifying assumptions on the model and approximations on the implementation of the estimation algorithm. In particular, the image is represented by a third-order MMRF and the proposed estimation algorithm is applied over the look-ahead observations rather than over the entire image. Thus, the formulation is derived from the extension of the look-ahead technique devised for a real-time decoding. Experimental results confirm that the proposed approach offers a great potential for solving difficult handwritten character recognition problems under reasonable modeling assumptions.