الفهرس | Only 14 pages are availabe for public view |
Abstract Both digital Image and video compression are important fields in image processing. In recent years, applications such as videoconferencing and video telephone have been used in a variety of fields. These applications require real time compression and decompression. Thus, the need for new image and video compression algorithms that run in real time is increasing. Fractal image and video compression has attracted a lot of interest because it provides high compression ratios, good image qualities and fast decoding. At high compression ratios, tractal image compression has been shown to give superior quality images. However, fractal¬based image encoder complexity is typically much higher than that of the decoder. This makes this method more suitable for publishing or broadcasting where the image is compressed once by a central processor and decompressed several times by smaller receiving processors. And in the same time limits the use of fractals in interactive applications, which require fast encoding and decoding, like videoconferencing. In this thesis, we suggest using block signatures and signature trees in fractal image and video encoding. This method is an extension for the feature vector approach suggested by Saupe, where the block feature vector is replaced with a signature for that block in a way that comparing the signatures of two blocks gives a good measure of the similarity between them. Two methods are used in creating block signatures, the pixel average and the Discrete Cosine Transform (DCT) methods. Also, the block classification suggested by Fisher is incorporated with the new method to achieve better performance. The new method is very suitable for parallelism where it is expected to reduce the compression time drastically to achieve real¬time applications requirements |