Context-based Still And Aurora Image Compression

With the development of digital communication and Internet, the demand of multimedia information, such as data、voice、image and video etc, is becoming larger and larger. At the same time, the expectation of multimedia quality is becoming higher and higher. Image, as an important component of multimedia, contains a great amount of data. So, mass storage and transmission of image impose high requirement on the technique of image compression.Wavelet transform has the properties of both time domain and frequency domain and compression technologies based on wavelet transform are able to achieve embedded bit stream, so, wavelet transform has found wide applications in the field of image compression. Based on the research of wavelet theory and context modeling, this paper improve traditional arithmetic coding based on context modeling and lossless compression based on context prediction. Further more, based on the theory of motion compensation, this paper further improves the prediction-based lossless compression algorithm. This study mainly includes following parts:(1) Image compression based on off-line and adaptive weight context classification. This paper proposes a new arithmetic coding based on both off-line and adaptive weight context classification against the problems of quantization complexity and context dilution in high-order context arithmetic coding. Through constructing context modeling using the correlations among wavelet coefficients, we can get the weight which represents the importance probability of current pixel/block. Regarding the weight as context and further reducing the context order through classifying the similar or close weights into one context using Lloyd-max algorithm, high-order context arithmetic coding can be approximated as low-order. This paper includes both on-line and adaptive weight context classification and weight probability estimation, and the performances both outperform state-of-art context-based arithmetic coding.(2) Aurora image compression based on context prediction. This paper proposes a novel adaptive prediction algorithm based on context modeling. The context modeling used in this paper is able to choose intra/inter-frame prediction adaptively based on the intra/inter-frame correlations of aurora image. At the same time, the proposed algorithm can predict current pixel adaptively using proper pixels in the context modeling and solve the disadvantages of using pixels with fixed position and fixed numbers. The proposed algorithm has low complexity and can satisfy real time application. Further more, it outperforms traditional lossless compression algorithms based on prediction or transformation.(3) Lossless compression of aurora image based on motion compensation. Based on motion compensation and the non-rigid specific character of aurora image, a prediction algorithm based on pixel motion estimation is introduced. The proposed algorithm overcomes the problem of inaccuracy results from the block-based motion estimation, and outperforms the performances of traditional prediction-based algorithm and block-based motion compensation. At the same time, the proposed algorithm has low complexity and can satisfy real time requirement.