干涉高光谱图像的主分-调制预测无损压缩

Lossless compression of hyper-spectral interference image based on principal-modulated prediction

提出一种基于主分调制预测的高光谱图像无损压缩算法。将干涉高光谱图像分为空间方向和光程差方向,空间方向采用主分量预测算法来去除帧间冗余;光程差方向采用调制分量预测算法来去除谱间冗余。主分量预测采用两步预测算法,第一步采用四阶预测器得到预测参考值,第二步采用8级查找表搜索预测算法得到实际预测值,然后将参考预测值和实际预测值进行比较得出最后的预测值。调制分量预测采用线性预测得到调制预测帧。最后,根据主分预测帧和调制预测帧得到最终预测帧,从而得出残差帧,利用残差帧进行熵编码。实验结果表明,文中算法的平均压缩码率达到3.05bpp,与传统高光谱图像无损压缩算法比较,平均压缩码率提高了0.14~2.94bpp,有效地提高了干涉高光谱图像无损压缩码率。

A lossless compression algorithm of hyper-spectral interference image based on principal- modulated prediction is proposed. Hyper-spectral interference images are divided into the space direction and the optical path difference（OPD） direction. In the space direction, a principal component prediction algorithm is used to reduce the inter-frame redundancies. And a modulated component prediction is used to reduce the spectral redundancies in the OPD direction. A two-step prediction algorithm is proposed for the principal component prediction. In the first step of prediction, a four order predictor is used to obtain a prediction reference value. In the second step,an 8-level lookup tables＇ prediction algorithm is proposed and used to obtain the real-prediction. Then the final prediction is obtained through comparison between the real value and the reference prediction. A linearity prediction is used to obtain modulation prediction frame in the modulated component prediction. Finally, the final prediction frame is obtained through comparison between the principal component frame and the modulated component prediction frame. And the residual frame is obtained,which is encoded by an entropy coder. The experiments results show that the average compression ratio of proposed compression algorithm is reached to 3. 05 bpp. Compared with traditional approaches,the proposed method can improve the average compression ratio by 0.14-2.94 bpp. They effectively improve the lossless compression ratio for hyper-spectral image lossless compression.