Tifinagh document segmentation based on Texture attributes

Text-graphic segmentation is a crucial step in document analysis pipeline, particularly for documents that contain a combination of textual content and graphical elements.This paper presents an approach for effectively segmenting text and graphic components in Tifinagh (alphabet used to write Amazigh language) documents.The proposed method consists of using the texture attribute to effectively detect and extract textual areas and graphical objects.Precisely, it is based on the design of two Gabor filter banks, where the first is configured with high frequencies to identify texts, and the second is designed to detect graphical components using low frequencies.