An Algorithm for Scanned Document Image Segmentation Based on Voronoi Diagram

With the emergence of complex layout, the layout isno longer confined to rectangular. This makes the traditional layout segmentation algorithm no longer applies, and new method dealing with complex layout emerge. This article proposed an algorithm for the scanned document image segmentation based on Voronoi diagram. The algorithm does not need additional tilt detection to oblique layout and pretreatment process for tilt correction, Voronoi diagram generated directly on the outer edge of the connected element, not related to the pixel processing, The algorithm does not need the processing to delete redundancy Voronoi edges and merger Voronoi edges, greatly reducing excessive segmentation of the other algorithm. The algorithm can fully identify each area of the document image, and the segmentation is accurate with very little information lost. In addition, because the structural elements were performed using statistical features, the adaptability of the algorithm is better.