基于疑似目标区域提取的空间图像降噪

Space image de-noising through getting the district where the targets possibly exist

对于空间目标图像的降噪预处理,要求能够完整地保留目标的边缘同时保证目标内部灰度分布不变,直接利用传统降噪增强方法难以取得理想效果。针对空间目标可见光图像,分析可能引入的噪声种类及其特性,提出了一种通过提取疑似目标区域达到降噪目的的方法。该方法首先使用SUSAN滤波结合频域分析、背景杂波抑制获得目标区域,然后从原始图像中提取出没有背景噪声的图像,从而实现图像的降噪。为了验证降噪效果,从整体视觉效果、局部星点视觉效果和信噪比三方面,与中值滤波、低通滤波、小波变换方法处理结果进行比较,实验结果显示,该方法在去除背景噪声的同时,能够更好地保留目标的边缘和内部灰度分布。

The image taken by the visible sensor carried by star is featured by little difference between target and noise, no atmospheric attenuation, and influenced by spherical radiation greatly. The de-noising to space images requires that the target edges be kept integrally while keeping its inner gray scale distribution unchanged. Traditional de-noising methods may greatly affect subsequent work of target recognition. Thus a method is proposed which reduces the noise through getting the district where the targets possibly exist in. The method makes use of SUSAN filtering, frequency domain analysis, and background clutter reduction to get the target district, then extracts the image that has no background noise and keeps original target brightness distribution from the original image, and thus the image de-noising is realized. To verify its de-noising effect, it is compared with median filtering, low-pass filtering, and wavelet transform from three aspects. The result shows that the method can keep the target edges and inner gray scale distribution well while reducing the background noise.