Characteristics Of Block-based Fractal Image Compression

Fractal image compression is one of the extensive schemes of digital imagecompression. It has received much attention from the research community dueto it's desirable properties such as new theoretical background, resolutionindependence and fast decoding. In this dissertation the fractal approximation ofcurves is first studied. According the characteristic of curve, a method regardingfiltering local disturbance is introduced. Based on the selection of characteristicpoints of the overall curve, the curve is simulated by fractal interpolationfunction. The experments for CT data curves and speech wave curves show thatthe method proposed in this dissertation is simple and efficient for curvessimulation.The characteristic points of curve are generalized to the characteristicblocks of image based on the studied of curves. An given image is partitionedwith it's characteristic blocks. Then it should be worked out image simulationbased on Coons fractal patches and the characteristic blocks-based fractal imagecompression. In the later, the regional shape of partitioned image is encoded andthen matched with other similar part of the image, which is the same as theoriginal region in size and shape. As a result, the contractive mappings can beachieved in order to make the fractal image compression automatically realized.It is tested that the approach we proposed is simple but available all through theexperiment. At the same time, a higher compression ratio could be obtainedunder the condition of maintaining the quality of the decoding image.