Efficient region merging method based on classified merging cost

During image segmentation, the hierarchy queue in the traditional region merging method has lots of redundant links, and it spends too much time on searching and sorting for keeping links in order. These have seriously affected the merging efficiency. Therefore a new region merging method based on quantified merging cost was proposed. It proposed several constraint conditions to shorten the queue length. Those links which should be put in the queue were quantified according to merging costs so that links with nearly the same cost were put in a same category, and their sorting were unnecessary and skipped. Furthermore, the MAP class in STL was utilized to build a two-dimensional dynamic queue, so that the time consuming will be reduced owing to its advantages in searching and auto-sorting. Comparative experiments showed that his method could not only guaranty the segmentation accuracy, but also obviously improve the merging efficiency.