Fractal Image Encoding In Wavelet Domain And Fast Implementation On GPU

With the development of information technology, image and video are some important kind of information carriers in our life. But the raw digital images are very big in size. If the image date couldn’t be compressed efficiently, it will not meet the needs of high speed transmission and efficient storage nowadays. Therefore, image compression, as a crucial and key technology in the fields of communication and multimedia, is rapidly expanding now and may have a promising future.The relatively low compression rate and poor quality of decoding image make the urgent need of new image compression methods. Fractal image compression pays attention to the excavation of the image self-similarity, using some simple iterative algorithms to coding a seemingly complex image. The decoding part can also use these algorithms to reconstruct images and to reach a very high compression rate. The wavelet transform can efficiently process the features of signals for its time-domain and frequency-domain. What’s more, wavelet transform can eliminate the overall correlation and makes the image be human vision characteristics. Therefore, the coding technology of fractal compression and wavelet transform are widely used in the fields of image compression. The excellent link of the two technologies not only represents the high compression rate of image compression, but also makes up the flaws of long coding time and blocking effect. CUDA, the GPU high performance computing architecture is explored in this paper to code the fractal image compression in wavelet domain, thus shortening the encoding time and increasing its efficiency. The main work of the dissertation is as follows:First of all, a brief description about image compression.Then study the fractal image compression. The result shows that this method reaches a very high compression rate based on a better image quality, but it needs longer time for coding and revels an evident blocking effect.Then in order to improve the above-mentioned algorithm, a further study about the fractal image compression in wavelet domain was made. The programming realized the fractal image compression in wavelet domain. The result shows that this kind link makes the fullest use of the two algorithms, which both shorten the coding time and avoid the blocking effect.Last in order to make a further step to promote the efficiency of image compression, this paper aims to use GPU parallel program to realize the algorithm.Under the CUDA specification, GPU was used to search the best match parent-tree for each sub-tree parallel. Experiment results proved that the scheme could magnificently reduce encoding time to milliseconds-level with maintaining the reconstructed quality, it may make the compression algorithm more efficiently.