基于地物反射特性的多波段遥感图像无损压缩方法

Lossless Compression of Multispectral Remote Sensing Image Based on Object Reflection

要实现多波段遥感图像的无损压缩，去除空间和谱间的相关性是其中的重要环节。通过分类，分析不同地物在不同波段的反射特性，再根据不同地物的波段反射特性构造出针对不同地物的谱间预测器，用以去除谱间相关性。对于去除谱间相关的残差图像，采用Ｓ＋ Ｐ变换，以去除空间相关。实验取得了令人满意的效果，证明了这种方法的有效性。

The spatial and spectral decorrelation are important steps to compress losslessly multispectral remote sensing image. At present, there are many methods to realize spatial and spectral decorrelation. But few of them is really effective. In this paper, we try to propose such an effective method in which the reflection properties of different objects in different bands were firstly analyzed using a wide-known classification scheme which called K-mean method. Then, for purpose of spectral decorrelation, the inter-band predictors corresponding to different objects were designed using the Central Band Vectors corresponding to different objects, which describe the reflection properties of different objects. As to spectral decorrelated difference image and the so-called Original Frame of multispectral remote sensing image, the S+P transform was adopted for spatial decorrelation. The satisfactory results which showed effect of this method were obtained .