Segmentation-based JPEG for document images

This paper describes a method for compressing document images using variable quantization tables based on document segmentation. This method aims at solving the problem of text and fine details degradation often encountered when compressing images, especially scanned document using the JPEG standard. The segmentation step first mark image elements as belonging to high/low frequency areas. Using this information, higher quantization is selected for low frequency areas while the quality of fine details is preserved by using lower quantization factors.