中文摘要：

边缘检测算法通过识别每个像素在其邻域内灰度的变化而检测出图像的边缘，但由于黄土CT图像的复杂性和随机性。经试验验证，典型的边缘检测算法并不适合直接应用于黄土微结构图像的处理.对边缘检测算法进行了改进并通过VC＋＋6．0平台实现了其功能。首先使用中值滤波和数学形态学方法对图像进行了处理，有效地降低了噪声干扰，净化了背景并增强了图像对比度；然后对图像进行基于Prewitt算法的边缘检测。通过试验验证了改进后的方法可以有效的去除虚假边缘。区分黄土颗粒的形态，从而更清晰地揭示其内部物质的结构。该方法对CT图像处理范畴的拓展，黄土微结构的研究方法和黄土工程性质的判定都有积极的指导意义。

英文摘要：

Edge detection algorithm could detect the edge of image thought identify gray-scale changes of each pixel in its neighborhood.However,due to the complexity and randomness of the loess, CT images.Verified by experiment,the typical edge detection algorithm is not suitable for applied directly to processing microstructure of loess image.This paper improves edge detection algorithm and achieve its function by VC 6.0 platform, First use the median filter and mathematical morphology process image,reduce the noise effectively,clean background and enhance the image contrast;Then detect image edge based on Prewitt algorithm.It verifies by testing that the improved method can remove the false edges effectively,distinguish the shape of loess particles and thus reveal the internal structure of matter more clearly.