基于预测与JPEG2000的高光谱图像无损压缩方法

Lossless compression of hyperspectral image based on prediction and JPEG2000

随着成像光谱仪向着高光谱分辨率、高空间分辨率方向发展,高光谱图像的数据量呈几何级数增长。由于数据传输和存储能力的限制,必须对高光谱图像进行有效压缩。首先,对高光谱图像的相关性进行了深入分析,得知其具有一定的空间相关性和极强的谱间相关性,从而具有较强的可压缩性。其次,结合JPEG2000对DPCM进行了修改,提出了基于一阶线性预测与JPEG2000相结合的无损压缩方案。最后,在软件平台上实现了该方案,并取得了较好的压缩效果。结果表明,该方案可以有效的实现高光谱图像无损压缩,验证了方案的可行性,为硬件平台上实现该方案提供了理论依据。

With the development of high spectral and spatial resolution imaging spectrometer,the amount of hyperspectral image data dramatically increase.Due to the limitation of transmission and storage ability,hyperspectral image must be compressed effectively.Firstly,the correlation of hyperspectral image is deeply analyzed,and it shows that the image has a certain spatial correlation and an extremely high spectral correlation.That makes it possible to compress the image to a great extent.Secondly,the scheme of lossless compression based on one-order linear prediction and JPEG2000 is proposed after the modification of DPCM combining with JPEG2000.Finally,the scheme is realized by programming and preferable results are obtained.The results show that hyperspectral image can be effectively compressed with the scheme,which verifies its feasibility and provides the theoretical basis for its implementation on the hardware platform.