Document image compression by subband system

With the advance of optical storage devices, a large amount of office documents can be stored in the form of scanned images which allow a more efficient retrieval system and ease of handling. Image compression is applied to increase the storage efficiency. Traditionally, a subband system has suffered from basis related visual artifacts. Since the document image contains mainly sharp characters and drawings on a monochrome background, strong ringing is expected as the image contains strong edges. The ringing effect is further worsened by the lack of a noise masking background. In this paper, a subband filter with nonnegative scaling functions and small mean square spatial width is considered to suppress the ringing effect and provide good spatial localization for sharp edge reconstruction. A quantization scheme is investigated to exploit the characteristics of the document image for more visually pleasing results. Design examples with high spatial localization property and their application in document image coding are presented