【摘要】 根据超光谱图像有很强谱间相关性的成像特点,设计了一种预测算法结合变换编码算法的图像分层压缩方法。采用几个相邻波段图像共享同一有损图像作为预测图像,克服了预测算法对传输误差的敏感性。预测图像是通过对原始图像进行局部熵为代价函数的四叉树分割后,再经小波变换得到的。用各个原始图像减去共享预测图像来去除谱间相关性,并得到相应残差图像,再对残差图像采用局部离散余弦基变换方法去除空间相关性,实现近无损压缩。研究结果表明,各波段峰值信噪比(PSNR)为40dB左右时,压缩比(CR)高于7.2,分层压缩方法具有很好的压缩效果。更多还原

【Abstract】 Based on the strong spectral correlation of a hyperspectral image,a multi-layered decomposition compression method combined prediction algorithm with transform encoding algorithm was proposed.Taking the same loss image shared for several neighbor bands as the prediction image,the sensitivity of transmission error for prediction algorithm was improved.The loss image was created by using quad-tree partitioning of original image and compressing the input image with a wavelet basis.After calculating the residual image between the original and loss image data,the rid of the spectral correlation was obtained,and a local cosine transform was used for removing the spatial correlation of residual image.The experimental results indicate that the compression ratio(CR) of the image is higher than 7.2 when the Peak Signal Noise Radio(PSNR) range is about 50 dB,which shows that the algorithm is efficient.更多还原