Hierarchical Image Segmentation based on Iterative Contraction and Merging

Abstract

In this paper, we propose a new framework for hierarchical image segmentation based on iterative contraction and merging (ICM). In the proposed framework, we treat the hierarchical image segmentation problem as a sequel of optimization problems, with each optimization process being realized by a contraction-and-merging process to identify and merge the most similar data pairs at the current resolution. At the beginning, we perform pixel-based contraction and merging to quickly combine image pixels into initial region-elements with visually indistinguishable intra-region color difference. After that, we iteratively perform region-based contraction and merging to group adjacent regions into larger ones to progressively form a segmentation dendrogram for hierarchical segmentation.

Existing System

Multiscale Normalized Cut (MNCut) method, graph-based image segmentation (GBIS) Algorithm, Normalized Cuts (NCut), Mean-shift algorithm (Mshift).

Proposed System

In this paper, we propose a rather different approach for hierarchical image segmentation. In the proposed approach, the progressive region merging is achieved by iteratively performing a contraction-and-merging process. Here, the contraction process is an optimization process based on an affinity matrix to condense image pixels/regions, while the merging process is a grid-based process to group pixels/regions that have been pulled sufficiently close. In the proposed framework, instead of extracting the global information in one step by spectral clustering, like that in the *-OWT-UCM algorithms, we progressively explore the global information of the image as the iterative process proceeds. Besides, the affinity matrix used by the contraction process is dynamically adjusted to update the color/size/texture/intertwining information for region merging. The proposed algorithm can automatically construct a segmentation dendrogram and the segmented regions are guaranteed to be consistent at different resolutions.

Architecture

BLOCK DIAGRAM