基于图像金字塔与NLM的冷冻电镜图像增强方法

Cryo-EM Image Enhancement Method Based on Image Pyramid and Non-local Means Denoising

针对冷冻电镜图像信噪比极低、颗粒与背景之间难以区分问题,提出了一种基于图像金字塔与非局部均值(NLM)去噪的冷冻电镜图像增强方法。先利用图像金字塔技术获取冷冻电镜图像中不同尺度的信息;然后在构建图像金字塔过程中循环使用NLM算法去除图像中的噪声,并运用图像背景矫正方法降低图像中的亮度不均现象;再对图像金字塔各级图像使用直方图变换方式进一步增强图像中颗粒与背景的区分度,最终使用图像上采样与图像融合技术将图像金字塔中的多层图像融合,形成冷冻电镜图像的增强图像。在原始与模拟的电镜图像上进行了实验,结果表明该方法能够有效提高图像峰值信噪比和结构相似度,颗粒与背景之间的区分度有显著提升,颗粒特征更为明显。

The problem that the extremely low signal-to-noise ratio of cryo-EM images and the unclear boundary between particles and background becomes an obstacle to particle selection,hence,an image enhancement method based on image pyramid and non-local means(NLM)denoising is proposed.Image pyramid technology is applied to obtain information at different scales in cryo-EM images.In the process of constructing the image pyramid,the NLM algorithm is used to remove the noise in the image,and the image background correction method is used to reduce the uneven brightness in the image.Then,histogram transformation is used to further enhance the distinction between particles and background.Finally,the image up-sampling and image fusion methods are used to fuse the multi-layer images in the image pyramid to form an enhanced image.The experimental results on real and simulated cryo-EM images show that the proposed method can effectively improve the peak signal-to-noise ratio,structural similarity of the images and the distinction between particles and background,and can obtain more distinct particle features.