Link Shifting Based Pyramid Segmentation for Elongated Regions

The goal of image segmentation is to partition an image into regions that are internally homogeneous and heterogeneous with respect to other neighbouring regions. An improvement to the regular pyramid image segmentation scheme is presented here which results in the correct segmentation of elongated and large regions. The improvement is in the way child nodes choose parents. Each child node considers the neighbours of its candidate parent, and the candidate parents of its neighbouring nodes in the same level alongside the standard candidate nodes in the layer above. We also modified a tie-breaking rule for selecting the parent node. It concentrates around single parent nodes when alternatives do not differ significantly. Images were traversed top to bottom, left to right in odd iterations, and bottom to top, right to left in even iterations, which improved the speed at which accurate segmentation was achieved. The new algorithm is tested on a set of images.