卫星干涉高光谱遥感图像序列差值压缩编码

根据干涉高光谱图像的成像特点和卫星遥感图像压缩编码系统要求,提出了一种新的卫星干涉高光谱图像序列压缩算法。新算法利用干涉成像光谱仪推扫成像特点提出了一种低存储量,帧间小波域匹配的图像序列压缩方案,提高图像质量3—4dB。算法首先通过小波域匹配算法检测出相邻图像之间的推扫位移量并移位求得两帧的最佳差值图像,然后对差值图像进行静止图像EBCOT压缩编码,从而提高总体编码效率。该算法具有与单帧图像编码算法相当的复杂度,只需要对当前图像与模板的差值图像进行基于小波变换的编码,避免了目前基于三维小波变换的编码算法对系统大存储量要求以及编码延时大的缺陷。仿真结果表明,本算法针对干涉高光谱图像编码,可获得比基于三维小波变换的算法相当甚至更好的效果,并且大大降低了编码延时。

THE COMPRESSION CODING OF SECONDARY PLANET INTERFERENTIAL SUPERSPECTRICAL IMAGE SERIES

A new second planet superspectrical image series compression coding algorithm is proposed in the paper. The algorithm adopts the still image compression method EBCOT （Embedded Block Coding with Optimized Truncation） as the coding engine, but it can efficiently reduce the redundancy among the frames. In the most documents on multispectral or superspectral images compression, the redundancy among the frames were usually reduced by 3 dimensional compression algorithm, such as 3 dimensional SPIHT, which in general based on 3 dimensional transform. Therefore, the huge memory is necessary when the transform is done, but the hardware system on the second planet must be lightest and with the lowest power consumption, and 3 dimensional compression algorithms are difficult to be used on the remote-sensing second planet. The novelty of the proposed compression coding method is that the interferential imaging characteristic of progressive scan is utilized to do the matching operation in the wavelet domain, when the relative shift between the two successive images is detected by the match method, the difference （error） image of the two images is encoded in stead of the origin image. Therefore the whole coding efficient is improved and the fidelity of the images can be improved by 3—4 dB. Since the complexity and memory is similar to the 2 dimensional compression algorithm, the algorithm is easier than the one based on 3 dimensional wavelet transform to be implemented on the second planet, and the disadvantage of large memory requirement and long encoding delay is overcome also. Numerical experiments on the 512-frame superspectrum image series show that the reconstructed images after using the proposed algorithm exhibit a comparative or even better effect at the same compression ratio than the 3 dimensional methods.