The authors have proposed a method for extracting characters from given images without prior knowledge of their position and size in the images. Based on an assumption that characters have line-type shapes locally and blob-type shapes globally, scale space analysis is used to detect both local and global shapes. In the scale shape analysis, given images are blurred by Gaussian filters at various scales. As the scale increases, images are more and more blurred, and more global shapes appear in the image. Their proposed method detects the scales at which global shapes appear by referring to normalized principal curvatures of the gray-level image profiles.
Character extraction method without prior knowledge on size and position information
2001-01-01
506805 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Character Extraction Method whithout Prior Knowledge on Size and Position Information [20014552]
British Library Conference Proceedings | 2001
|Hierarchical pre-segmentation without prior knowledge
IEEE | 2001
|Position Finding with Empirical Prior Knowledge
IEEE | 1972
|Hierarchical Pre-Segmentation without Prior Knowledge
British Library Conference Proceedings | 2001
|