Multidimensional Image Lossless And Lossy Compression Based On The Context

The vast data of multidimensional image bring the great problems in the storage and data transmission. Therefore, according to the feature of multidimensional image, researching the high-efficiency compression technique of multidimensional image is significant for information transmission and storage.In our paper, we first introduce the compression background of the hyprespectral image at present; utilize the high correlation of neighbour bands, we proposed the lossless/lossy compression for hyperspectral image based on the context。Then based on the tranditional arithmetic coding, the paper proposed the High Dynamic Range arithmetic coding based on context, and using the different prediction methods for32bits Aurora image lossless coding. In short, as followed there have three main points in our method.1. At first, we take the non-linear2D-CALIC context prediction in intra-frame, and decompose the obtained the predicted error into bit-planes, then coding the bitplanes using LDPCA(Accumulated Low-Density-Parity-Codes) combined the DSC (Distributed Source Coding) theory. We analyse the results of non-uniform and uniform source coding with LDPCA. In this point, we realize the lossless compression of hyperspectral and improve the performance.2. We provide the principle of distributed arithmetic coding (DAC) and decoding based on the tranditional binary arithmetic coding. And give the SPECK coding theory under the wavelet transform, then apply the distributed arithmetic coding to the sign bitplane of SPECK coding, from the results we can see that the coding performance is preferably.3. While the arithmetic coding based on context is superior to the binary arithmetic coding, in our paper, we proposed the HDR (High Dynamic Range) coding of arithmetic coding based on the context modeling, we utilize the high correlation of neighbour bands to achieve the prediction by efficient LUT(Lookup table), then coding the predicted error by HDR arithmetic coding. Because of the bandwidth and low bit-rate requirements, we should realize the near-lossless compression of hyperspectral image.4. Due to the tremendous data of three-dimension Aurora images obtained in China’s Arctic Yellow River station, the long distance transmission and efficient compression become a key.The paper introduce several efficient prediction method for three-dimension image, and appliy the prediction methods for three-dimention Aurora images lossless coding.