摘要： 提出了一种基于形态学多尺度的图像噪声去除方法，该方法首先利用形态学多尺度开闭重建运算对噪声图像进行多尺度重建，将噪声图像分解为一系列尺度不同的特征图像叠加，然后对叠加特征图像进行尺度模式谱分析，确定图像中噪声对应的尺度范围，最后将噪声尺度对应的特征图像从噪声中去除，达到同时消除噪声和保持图像目标信息完整及准确定位的目的。仿真实验表明，该方法能够有效地去除不同类型的图像噪声，具有较高的输出信噪比，同时保持了图像信息的完整和图像目标的准确定位。

关键词: 噪声去除；形态学多尺度；开闭重建运算；模式谱；信噪比

Abstract: A morphological multi-scale approach for image noise removing is proposed. Morphological multi-scale opening and closing by reconstruction are employed to noise image, which decomposes the noise image into different scaled feature image; scale pattern spectrum analysis is used to confirm the scope of scale corresponding to noise; and noise scaled feature images are removed from the original noise image. Experiments show that this method can efficiently not only remove various typed noise with high SNR, but also hold the integrity of image content with less shape changing and the precision of objects localization without evident bias.

Key words: Noise removing; Morphological multi-scale; Opening and closing by reconstruction; Pattern spectrum; Signal-noise-ratio(SNR)