Text segmentation in vintage floor plans and maps using visual perception

Abstract

Floor plans and maps are types of graphical documents which have similar features as both of them have their information described through text and drawings for example. This paper presents a new algorithm for text segmentation in images of vintage floor plans and topographic maps. The algorithm is divided into: background removal, thresholding, histogram equalization, connected component analysis, lines removal, noise removal, and restoration. The experiments were conducted in a dataset of real old floor plans and maps from different ages and the results are faster and better than other algorithms.