摘要
针对压缩时不能充分利用图像普相关性、细节信息丢失等问题,提出一种自适应分组Karhun-Loeve变换/整数小波变换(APKLT/IWT)的高光谱图像压缩方法。对高相关性的谱段进行自适应分组,应用改进式KL变换和整数小波变换分别消除谱相关性和图像空间冗余,应用SPIHT算法进行压缩编码,对低相关性的谱段则直接采用三维整数小波变换/三维SPIHT(3DIWT/3DSPIHT)进行压缩编码。实验结果表明,相同压缩比条件下,PSNR值比非自适应的分组KL变换/整数小波变换/SPIHT(KLT/IWT/SPIHT)压缩算法提高了6dB,运算时间减少,压缩性能明显提高。
Because the compression can not take full advantage of the image of the general relevance,and losses information details and other issues,a compression method for hyperspectral images based on adaptive partitioned Karhun-Loeve transform（APKLT）and integer wavelet transform（IWT）was proposed.The spectral data of high correlation were partitioned into several classes.The spectral correlation and image space redundancy of these classes were eliminated by KLT and IWT respectively.The SPIHT algorithm was used to compress the APKLT/IWT data.The spectral data of low correlation were compressed using3 DIWT and 3 DSPIHT algorithm directly.Experimental results show that the PSNR of proposed APKLT/IWT/SPIHT method improves 6 dB comparing with non-adaptive KLT/IWT/SPIHT method in the condition of same compression rate,and the computing time decreases,which outperforms other methods.
出处
《计算机工程与设计》
北大核心
2018年第3期780-784,共5页
Computer Engineering and Design
基金
国家自然科学基金项目(61672335)
