基于颗粒分析的Top-Hat变换及去噪

AN IMPROVED ALGORITHM OF TOP-HAT TRANSFORM BASED ON GRANULOMETRY AND NOISE REDUCTION

数学形态学中的Top-Hat变换是一种提取图像中局部较亮的目标或信息的技术。经典的Top-Hat变换用唯一尺寸的结构元素对图像进行处理,并不能有效地提取目标,图像上面的结构或目标不一定对那个尺寸的结构元素敏感。通过定义一个递增的结构元素序列对图像进行Top-Hat变换,在每一个像素处用不同的结构元素作Top-Hat变换,以寻找对该像素最敏感的结构元素,并把此结构元素对图像的处理结果作为该像素处最后的输出值。虽然该方法有效地增强了图像,但同时也增强了噪声,最后利用形态重构算子对噪声进行了处理。

This paper proposed an improved algorithm of Top-hat transform based on graunulometry. Top-Hat transform can extract lighter objects and structures in gray image. Classical Top-Hat transform only uses structure element in one size and can not effectively extract objects in the image, because only part of objects may have a high response for a given structure element size and a lower response for the rest. In this paper, an increasing sequence of structure element was used to detect the best response to each pixel of image and the output of the response pixel would be the most sensitive structure element by using this improved Top-Hat transform. This method improve the image but improve noises too. So morphological reconstruction is used to denoise at last.