Orientation and Scale Invariant Text Region Extraction in WWW Images

Text extraction from a web image is important for web indexing because the text can contain a key information of the web. This paper presents a method to detect a text with various orientation and multifont sizes in a web image. The proposed method consists of three steps; 1)color reduction of low resolution web image by a spatial merging algorithm, 2) extraction of character candidates by finding connected components corresponding to strokes, and 3) detection of texts in an arbitrary direction by filtering and combining character candidates based on a potential field. This method allows to detect text orientation and scale invariantly. Experiments with 200 web images are performed to verify the validity of the proposed method.

• Type: Conference
• Year: 1998
• Where: MVA
• Authors: Taehoon Park, Dongsung Kim, Kyusik Chung
• Keywords: Character Candidates, MVA 1998, MVA 2007, Text Extraction, Web Images