An efficient strategy for implementing iterative area openings using the max tree

Area opening is an important morphological connected set operator that features in removing upper level sets from an image whose area properties are smaller than a threshold lambda. Existing algorithms found in the literature that implement the area opening operator are based on either priority queues, the max tree or the union-find approach. In this paper we explore the advantages of using the max tree based approach for iterative area opening. Iteratively applying an area opening is the central idea underpinning all scale based image decompositions. An efficient implementation strategy for iterative area opening is therefore very important if scale based image processing algorithms are to be successfully applied in real time computer vision applications. This paper builds on recently published work comparing approaches for implementing area openings, and improves on the method proposed for image reconstruction via a max tree. Experimental results are presented that show the new approach proposed in this paper obtains a performance gain of 25% with images of reasonable big size (320 × 256)