扫描电镜能直观观察样品的表面结构, 但其高分辨形貌成像图固有的噪声不利于图像分析。针对集成电路器件扫描电镜成像图的去噪声问题, 采用了通过滑动条方式自适应设置图像二值化阈值, 将数学形态学处理方法与图像二值化相结合, 实现了对图像噪声的自动去除处理; 同时还设计了通过手动勾勒图像中的多边形区域实现去除噪声的功能; 为使图像达到更好的效果, 系统还可允许针对自动去噪后的图像自行选择是否进行手动去噪, 并设计实现了风格直观简洁, 易于操作的交互式用户界面。对多幅集成电路器件扫描电镜成像图进行去噪声处理的结果和对去噪前后的图像进行无参考图像质量评价的数据表明, 该方法有效地改善了扫描电镜图的信噪比, 获得了突出前景等有用信息。

The scanning Electron Microscopy(SEM) can observe the surface structure of the sample directly. However, the noises remaining in its high-resolution image will affect the image analysis. Addressing to the problem of SEM image denoising, the automatic and manual image denoising function is achieved. In the image denoising algorithm, both mathematical morphology and binarization are adopted. Moreover, the image binarization threshold can be adaptively realized by adjusting a designed slider and the images after automatically denoising can be further reduced its local noise by manual denoising process. A concise and easy graphical user interface(GUI) is presented. The experiment and analysis of the SEM images denoising show that the signal-to-noise ratio of the images is improved effectively. The proposed technique has been applied to the SEM image denoising for micron or nanometer level semiconductor devices and obtained satisfactory results.