Detection of artificial and scene text in images and video frames

摘要: Textual information in images and video frames constitutes a valuable source of high-level semantics for multimedia indexing retrieval systems. Text detection is the most crucial step text extraction system although it has been extensively studied past decade still, does not exist generic architecture that would work artificial scene content. In this paper we propose both frames. The based on machine learning stage which uses an Random Forest classifier highly discriminative feature set produced by using new texture operator called Multilevel Adaptive Color edge Local Binary Pattern (MACeLBP). MACeLBP describes spatial distribution color edges multiple adaptive levels contrast. Then, gradient-based algorithm applied to achieve distinction among lines as well refinement localization lines. whole situated multiresolution framework invariance scale Finally, optional connected-component segments into words distances between resulting components. experimental results are applying concise evaluation methodology prove superior performance achieved proposed