Development of a semi-automatic open source tool for layout analysis and region extraction and region classificiation (LAREX) of early prints.

该提案的目标是进一步开发我们的高效、半自动和易于使用的开源分割工具LAREX，并将其集成到OCR-D功能模型的开源工作流中。LAREX(布局分析和区域提取)的初步工作既允许通过分离文本和非文本进行粗切分，也允许通过检测和分类不同的文本实体进行精细切分。LAREX利用连接组件方法的高效实现。它已被用于不同早期印刷品的数字化，并且能够以比传统替代方案显著更少的时间实现质量良好的页面分割。LAREX进一步发展的主要目标是降低人工劳动的程度。因此，更稳健的分割以及规则和约束语言的进一步发展是必要的。基本配置应该容易地适应用户和学习算法的特定早期印刷品的特点。此外，LAREX用于校正单个分割错误的舒适的图形用户界面应该得到改进。这一组成部分对于定义学习算法和评估的基本事实也是必要的。总体目标是在手动和自动方法之间找到最佳组合。将与各种合作伙伴一起对该工具和工艺模型进行实质性评价，特别是在OCR-D功能模型内的早期印刷品数字化的背景下，包括通过外部工具的联动实现随后的OCR。

The goal of the proposal is the further development of our efficient, semi-automatic and easy-to-use open-source segmentation tool LAREX und its integration in the open source workflow of the OCR-D functional model. The preliminary work LAREX (Layout Analysis and Region EXtraction) allows both a coarse segmentation by separation of text and non-text and a fine segmentation by detection and classification of different textual entites. LAREX utilizes an efficient implementation of the connected component approach. It has been used in the digitalization of different early prints und enables a qualitative good page segmentation with significantly less time than conventional alternatives. The main goal of the further development of LAREX is to reduce the degree of manual work. Therefore, a more robust segmentation und a further development of the rule and constraint language are necessary. The basic configurations should be easily adaptable to the peculiarities of a particular early print by both the users and learning algorithms. Furthermore, the comfortable GUI of LAREX for correction of single segmentation errors should be improved. This component is also necessary for defining a ground truth for learning algorithms and for evaluation. The overall goal is to find an optimal combination between manual and automatic methods. The tool and the process model will be substantially evaluated with various cooperation partners, in particular in the context of the digitalization of early prints within the OCR-D function model including the subsequent OCR by the linkage of external tools.