Multi-component document image coding using regions-of-interest

Region-of-Interest (ROI) techniques are often utilized in natural still-image coding standards such as JPEG2000 [1]. In contrast, document image coding typically adopts multi-layer methods [2], using a carefully selected algorithm for each layer to optimize overall performance. In this paper, an ROI-based method is proposed for multi-component document image coding, where rectangular textual ROI's are easily extracted using standard document image analysis techniques. Compared to multi-layer methods, the method is simpler and scalable, while preserving comparable visual quality at equivalent PSNR.