Robust image binarization with ensembles of thresholding algorithms

Abstract

The effectiveness of a thresholding algorithm strongly depends on the image statistical characteristics. In a completely unsupervised context, this makes it difficult to choose the most appropriate algorithm to binarize a given image. This issue is considered through a novel thresholding strategy based on the fusion of an ensemble of different thresholding algorithms and formulated within a Markov random field (MRF) framework. The obtained experimental results suggest that in general the fusion of an ensemble of thresholding algorithms leads to a robust thresholding system, and in particular the proposed MRF strategy represents an effective solution to carry out the fusion process.