Fuzzy Rule and Bayesian Network Based Line Interpolation for Video Deinterlacing

Gwanggil JEON  Jechang JEONG  

IEICE TRANSACTIONS on Communications   Vol.E90-B   No.6   pp.1495-1507
Publication Date: 2007/06/01
Online ISSN: 1745-1345
DOI: 10.1093/ietcom/e90-b.6.1495
Print ISSN: 0916-8516
Type of Manuscript: PAPER
Category: Multimedia Systems for Communications
deinterlacing,  fuzzy reasoning,  directional interpolation,  rough set theory,  Bayesian network,  

Full Text: PDF>>
Buy this Article

Detecting edge directions and estimating the exact value of a missing line are currently active research areas in deinterlacing processing. This paper proposes a spatial domain fuzzy rule that is based on an interpolation algorithm, which is suitable to the region with high motion or scene change. The algorithm utilizes fuzzy theory to find the most accurate edge direction with which to interpolate missing pixels. The proposed fuzzy direction oriented interpolator operates by identifying small pixel variations in seven orientations (0°, 45°, -45°, 63°, -63°, 72°, and -72°), while using rules to infer the edge direction. The Bayesian network model selects the most suitable deinterlacing method among three deinterlacing methods and it successively builds approximations of the deinterlaced sequence, by evaluating three methods in each condition. Detection and interpolation results are presented. Experimental results show that the proposed algorithm provides a significant improvement over other existing deinterlacing methods. The proposed algorithm is not only for speed, but also effective for reducing deinterlacing artifacts.