基于三维小波变换和上下文量化的高光谱图像压缩

Compression of hyperspectral images based on three-dimensional wavelet transform and context quantization

提出了一种基于三维小波变换和上下文量化的高光谱图像压缩算法.该算法先利用三维小波变换去除高光谱图像在谱内和谱间的冗余信息,然后建立了一个高维时、频域的上下文预测模型以分析小波系数之间的相关性,并通过对增加条件熵的理论分析,合理地对该模型中的上下文进行量化,得到合适的编码上下文用于自适应的算术编码.试验结果表明,该算法不仅明显优于基于二维小波变换的JPEG2000标准,而且相对于同样基于三维小波变换的三维分层树集合分裂算法也有较大的提高.

A novel hyperspectral image coding method based on three-dimensional wavelet transform （3DWT） and context quantization was proposed. With this method, 3DWT was firstly adopted to remove the intra- and inter-spectral redundancy of hyperspectral images. Then the correlation of wavelet coefficients was analyzed by a high dimension context predicting model. After the theoretical analysis on conditional entropy, this context model was quantized to provide the coding contexts for the adaptive arithmetic coder to yield an embedded bit-stream. The experiments show that the proposed method is not only much better than the 2DWT-based JPEG2000 but also provides improvement relative to 3DWT-based three-dimensional set partitioning in hierarchical trees （3DSPIHT）.