基于分类和曲线拟合的干涉超光谱图像压缩

根据干涉超光谱图像的特点,提出了一种基于图像分类与曲线拟合的干涉超光谱图像数据分解算法,结合内嵌比特平面编码技术实现干涉超光谱图像的压缩。与JPEG2000一样,该算法实现了有损、无损压缩的兼容。将干涉超光谱图像数据分为主干涉区域与非主干涉区域两类,针对主干涉区域提出了一种相似匹配算法,而对非主干涉区域采用经验模式分解和二次曲线拟合方法进行数据分析,两种分析算法结合起来能够有效地对谱线数据进行分解,从而有利于取得更好的压缩效果。仿真结果表明,提出的算法可以使无损压缩的输出码率降低0.2~0.4bit/pixel,而近无损、限失真压缩的重建图像质量相应提高。

Interference Hyper-Spectral Images Compression Based on Classification and Curve-Fitting

A data decomposition algorithm for interference hyper-spectral images based on classification and curvefitting is proposed, by studying features of hyper-spectral images. Compression of interference hyper-spectral images is realized by combining the embedded bit-plane coding technology, which implements loss and lossless compression in the same algorithm just as in JPEG2000. The data of a spectral line are decomposed into two classes, main- interference class and non-main-interference class. And a similarity-based match method is presented for the data of main-interference class, while the data of non-main-interference class is processed by empirical mode decomposition and second-order curve-fitting algorithm. The data of a spectral line can be approached appropriately by combining the two analytical algorithms, which benefits lossless image compression. The simulation results show that the output rate is decreased by 0.2 - 0.4 bit per pixel for lossless compression, and the reconstructed image quality is also improved.