Historical documents are invaluable resources for understanding the development of civilizations and cultures. However, the transcription process of these documents comprises many challenges such as complex layouts, degradation, various handwritten styles, and skewed text. This paper presents an unsupervised approach for text line extraction in the Belfort Civil Registers of Births, a historical dataset containing a mix of printed and handwritten text with marginal annotations. The proposed method employs a series of image processing techniques to identify text line cores. The method also utilizes a dynamic gap identification and segment point localization strategy based on text density and histogram analysis to effectively identify the borders of the text lines in polygon shape. An XML file generation tool is then utilized to structure the resulting components and link them with their corresponding text. The method exhibits competitive accuracy in segmenting text lines on both the Belfort dataset and standard benchmarks such as the Saint Gall and READ Bozen datasets. This work contributes to the preservation and accessibility of historical documents by facilitating accurate transcription and structured data representation.