作者: BERRIN A. YANIKOGLU , LUC VINCENT
DOI: 10.1016/S0031-3203(97)00137-4
关键词: Segmentation 、 Ground truth 、 Document processing 、 Computer vision 、 Optical character recognition 、 Artificial intelligence 、 Set (abstract data type) 、 Pattern recognition (psychology) 、 Scale-space segmentation 、 Computer science
摘要: We describe a new approach for the automatic evaluation of document page segmentation algorithms. Unlike techniques that rely on OCR output, our method is region-based: quality assessed by comparing described as set regions, to corresponding ground-truth. Error maps are used keep track all errors associated with each pixel, regardless complexity. Misclassifications, splitting, and merging regions among detected system. Each error can be weighted individually system customized benchmark virtually any type task.