The noise generated by Charge-Coupled Device (CCD) will limit transmission efficiency of the remote sensing im age data transmitted from the satellite to the earth. Therefore, in order to efficiently suppress CCD noise and increase lossless compression ratio of the remote sensing images,an anisotropic diffusion method based on estimation of noise standard deviation is proposed. Firstly, based on the results of image homogeneity measurement and edge extraction, an automatic estimation method of noise standard deviation is proposed. Then, empirically linear regression model between noise standard deviation and gradient threshold of the conduction coefficient in conventional anisotropic diffusion is extracted using simulated noisy images. Based on results of noise estimation and homogeneity measurement, a noise-driven anisotropic diffusion method , which could adaptively modify conduction coefficient value and iteration number,is finally proposed. Also, Beijing-1 multispectral images and standard test images are used to assess noise estimation and noise suppression methods. The experimental results show that compared with other existing anisotropic diffusion methods,the proposed method not only achieves a better quality of image restoration but also efficiently increases lossless compression ratio of multispectral images. Furthermore, the proposed method will provide a novel thought for future development of onboard data processing system.%CCD在成像过程中所产生的噪声会限制星地间遥感图像数据的传输效率.因此,为有效抑制CCD噪声并提高遥感图像的无损压缩比,本文提出了一种基于图像噪声标准差估计的各向异性扩散方法.首先,结合图像同质性测度和边缘提取结果提出一种遥感图像噪声标准差的自动化估计方法.而后,通过噪声模拟的方式拟合出图像噪声标准差与经典式各向异性扩散方程传导系数中梯度阈值的线性关系.基于图像噪声标准差估计和同质性测度结果,最终提出一种可自适应性调整传导系数值和迭代次数的噪声驱动式各向异性扩散方法.本文中分别采用北京一号小卫星多光谱遥感图像和标准测试图像对所提出的噪声估计方法和噪声抑制方法进行评价.试验结果表明,对比其他各向异性扩散方法,本文所提出的方法不仅可实现更好的图像复原效果,还可有效地提高多光谱遥感图像的无损压缩比,并为今后星上数据处理系统的研究提供一种较新的思路.
展开▼
