Abstract: It may be difficult to extract distinctive features pertinent to a specimen when dealing with very low-contrast and low SNR cryo-Electron micrograph(Cryo-EM) images. A DoG(difference of Gaussian) based method to extract shape features was provided in this paper,which is aimed to get a binary image by using DoG. This binary image was used as an object mask and shape features were calculated for the largest object in the mask. The experiments showed this method has good results.

Key words: Particle picking, Macromolecular 3D reconstruction, DoG, Shape feature, Feature extraction