A new region-based segmentation method for complex document image analysis

Abstract

In this paper, a new region-based segmentation method is proposed for resolving issues associated with the complexity of backgrounds of complex document images. The proposed method processes the document image regionally and adaptively according to local features using two stages. First, an automatic localised multilevel thresholding method is utilised to recursively segment a specified block region into several layered image sub-blocks. Then the multi-layer region-based clustering is performed to aggregate layered image sub-blocks with homogeneous features into associated object layers. Experiment results on text extraction from complex document images demonstrate the effectiveness of the proposed method.