根据遥感图像局部相关性较弱、纹理复杂丰富的特点,提出了基于小波分析理论的自适应标量、矢量混合量化压缩方法。该方法根据遥感图像小波变换后高频子图的局部块纹理强弱将这些块划分为4类,对平坦块进行高倍压缩,对纹理块进行高保真压缩,使各块的恢复误差大致平衡。其主要特点是避免了矢量编码过程中的码书训练和码书搜索,因而时间性能好,并且对单幅图像的压缩比和峰值信噪比(PSNR)优于JPEG方法。此方法与K L变换去波段相关技术相结合,应用于多波段遥感图像压缩领域,收到了良好的效果。

With the increasing volumes of remote sensing data, data compression is receiving more and more attention. Adapting to the specialities of remote sensing data the low local correlation and the rich complex texture information, this paper presents an adaptive scalar vector hybrid quantization method for compression based on wavelet transform. According to textural intensity of every block in wavelet transformed high frequency subimage, we classify them into four classes. Compressing the plain block is at high compression ratio, and the textural block at high fidelity, The method enable the balance of the restore error of every block. This method is time efficient by avoiding the codebook training and searching, while surpass the performence of JPEG for single image. By combining with K L transformation, which is a kind of correlation reduction methods, we apply the presented method to multi band remote sensing image, and good results have been obtained.