| This thesis discusses algorithms and real-time architectures to implement a novel image compression scheme.; Image compression is used to reduce the amount of data needed to store and transmit images. Real-time techniques compress images "as fast" as they are generated.; Our image compression scheme employs a visual criterion to determine when a given area in an image is visible over its background. It uses simple integer computations, and so is amenable to real-time use with relatively simple hardware.; Two computational strategies have been developed: a quad tree based algorithm and a region growing based algorithm. In the first method, the quad tree data structure is used for systematic traversal of the image. Real-time architectures for this method have been developed, both for data in quad tree order, and in scan line order. An architecture for a modified scheme using surround information has also been designed.; Blocking artifacts noticeable in the quad tree based algorithm are reduced in our second computational strategy, which uses the region growing algorithm. Visually homogeneous areas are grown from starting pixels. Architectures for this algorithm have been presented.; To smooth out the blocking artifacts created both by quad tree traversal, and by tiling the image into independent sub-images, a postfiltering scheme has been developed, along with real-time architectures for its implementation.; A lossless coding scheme to encode pruned tree structures in the quad tree based algorithm has been designed, together with real-time implementations. Loss-less coding for the region growing algorithm has been discussed.; We have extended the work to color images. Combined coding of the three primary color components, as well as separate coding, have been investigated.; Finally, the compression of image sequences has been studied. The tuning of compression ratio to channel capacity has been investigated for sequences compressed frame-by-frame. An interframe coding algorithm to exploit temporal redundancy has been developed. We have presented a lossless coding scheme for the algorithm, and an architecture for its real-time implementation.; Important parts of this work have been realized in custom-designed integrated circuits. The algorithms have been implemented in software and tested on a variety of display devices. |
