基于视觉模型和图像特征的遥感图像压缩

遥感图像具有纹理复杂、边缘丰富的特点,通常难以实现高压缩比.统计了图像经过小波变换后在不同方向、不同级别子频带的能量分布.然后基于人类视觉系统(HVS)模型中的对比敏感度函数(CSF)和遥感图像的能量分布特征,提出一种不同频带小波系数的变步长量化方法,并把该方法应用于嵌入式小波图像编码器.实验结果表明,与不采用心理视觉量化的压缩方法相比,在相同压缩比下,峰值信噪比略为降低,但恢复图像的主观质量有一定程度的改善.

Remote sensing image compression based on visual model and image feature

With characteristics of complicated texture and rich edges, remote sensing image is usually hard to achieve high compression ratio. Energy distribution among wavelet subbands at different orientations and scales was statisticed, then an adaptive quantization method for different wavelet subbands coefficients was proposed, in the method contrast sensitivity function(CSF) of human vision system(HVS) and energy distribution characteristic were both considered. The quantization method was applied to an existing embedded wavelet image coder and the compression results show that comparing with method without psychovisual quantization, this method gets better visual quality under the same compression ratio (only with little peak signal noise ratio(PSNR) degradation).