OPTIMAL SET-PARTITIONING BASED ON GROUP QUALITY LIKELIHOOD USING PARTITION-GROWING ALGORITHM

Abstract

We consider the problem of partitioning a set of elements X into a minimum of m blocks to maximize the probability that each block constitutes a well-grouped subset of X. We employ certain characteristics of the Group Quality Likelihood to limit our search for an optimal partition within a smaller set of partitions. We develop a novel Partition-Growing algorithm based on the Branch and Bound strategy that produces the optimal partition in an efficient manner. We demonstrate how the algorithm can be applied to recognize Broken Characters in degraded document images. Experimental results conducted on Thai historical documents are very promising.