Line-based image coding using adaptive prediction filters

This paper proposes a line-based approach for image coding. Unlike conventional block-based image coding, the proposed method takes one line as a basic unit in coding and reconstruction. Such a structure allows us to perform better predictions between neighboring units and affords us more flexibility when coding each unit. We also present an efficient prediction method using adaptive prediction filters for this structure. Based on the neighboring information, we apply the best prediction filter on each line to get high-quality predictions. We show that the proposed line-based coding can perform comparatively or even better than the state-of-the-art block-based coding. When combing line-based and block-based coding together, the state-of-the-art image coding can be much improved.