Research On Fast Fractal Image Coding

The novel thought of fractal image coding technology broken through the theoretical framework of traditional coding methods. This technology inaugurates a new approach for image coding and becomes one of the up-to-date methods in image compression field, due to its high potential compression ratio which can get as high as 10000:1 for some special image, short coding time, the decoding feature of independence of resolution, favorable reconstructed image quality and so on. The main drawback of fractal image encoding is the expensive computational cost in the encoding process, although it has many advantages. Therefore, this author studies mainly how to reduce runtime in the encoding process while maintaining the reconstructed image quality. The main research work can be organized in the following two aspects:1. This thesis mainly researches on basic algorithm of fractal image coding (Jacquin algorithm), and summarizes improved politics of Jacquin algorithm; Moreover, for each improved politic, some specific algorithms are presented and comparison is given between each algorithm and another algorithms. This thesis mainly presents quadtree algorithm and shape feature algorithm, and verifies the validity of these algorithms through experiments.2. For the lack of fractal image algorithm's long encoding times, by analyzing the distribution properties of image block, a new feature, named eigenangle, is defined in this paper. Without considering isometric transformation, this feature can indicate the distribution properties of image block, and is the necessary condition of image block similarity. In this paper, by integrating eigenangle and feature vector strategy, a fast fractal image encoding algorithm based on eigenangle is proposed. This algorithm, on the premise of guaranteeing the decoding image quality, can greatly enhance the encoding times, and will powerfully impetus the practicality of the fractal image encoding.Lastly, the research progress of fractal image coding is viewed in the end of thesis.