Hierarchical Image Coding

Abstract

Besides of the well known techniques for image source coding, the concepts of hierarchical coding have become more and more popular. The basic idea of these coding schemes is not only to compress images, but additionally to give the compressed image data a destinct structure. The pixel matrix is decomposed or transformed into a hierarchically ordered set of separated data blocks, each of which represents a meaningful part of the total image information (e.g.different parts of spatial resolution). These blocks can be stored, accessed and transmitted separately and sequentially so that a partial or progressive image reconstruction becomes possible.

One of the major application areas of hierarchical coding techniques are pictorial information systems with access to large image data bases, where the separation into meaningful data blocks turned out to be a powerful method for a user- or device-adapted way of reducing the transmission of irrelevant image information. Hierarchical coding schemes allow the userfriendly design of image workstations, with features such as very quick retrieval of survey images, browsing through pictorial data bases and progressive transmission with continuously increasing spatial resolution. Time periods for image retrieval can be objectively and subjectively shortened. A hierarchical separation of image data blocks supports the compatibility and integration of different image terminal devices in distributed pictorial information system.

Some of the introduced methods of hierarchical decomposition of still images are reviewed with emphasis on hierarchical transform coding techniques, i.e. block-transform and S-transform coding. The latter will be described as a hierarchical image coding scheme which is particularly suited for large pictorial information systems. It allows real-time operation, supports a VLSI-friendly implementation and can be used both for reversible or irreversible image data compression.