Zhang Qin, Bajaj Chandrajit L
Institute for Computational Engineering and Sciences, University of Texas, Austin, TX 78712, U. S. A.
Far East J Appl Math. 2010 Aug;45(2):83-161.
The energy functional used in digitalized total variation method is expanded to a general form and a generalized digitized total variation (GDTV) denoising method is obtained. We further expand this method from 2-dimensional (2D) image to 3-dimensional (3D) image processing field. Cryo-electron microscopy (cryo EM) and single particle reconstruction are becoming part of standard collection of structural techniques used for studying macromolecular assemblies. Howerver, the 3D data obtained suffers greatly from noise and degradation for the low dose electron radiation. Thus, image enhancement and noise reduction are theoretically necessary to improve the data for the subsequent segmentation and/or structure skeletonization. Although there are several methods to tackle this problem, we are pleased to find that GDTV method is very efficient and can achieve better performance. Comparative experiments are carried out to verify the enhancement achieved by the GDTV method.
将数字化全变差方法中使用的能量泛函扩展为一般形式，得到了一种广义数字化全变差（GDTV）去噪方法。我们进一步将该方法从二维（2D）图像处理领域扩展到三维（3D）图像处理领域。冷冻电子显微镜（cryo EM）和单颗粒重建正成为用于研究大分子组装体的标准结构技术集合的一部分。然而，由于低剂量电子辐射，所获得的三维数据受到噪声和退化的严重影响。因此，从理论上讲，图像增强和降噪对于改善后续分割和/或结构骨架化的数据是必要的。虽然有几种方法可以解决这个问题，但我们很高兴地发现GDTV方法非常有效，并且可以取得更好的性能。进行了对比实验以验证GDTV方法所实现的增强效果。
