无需设定阈值的图像边缘检测

Image edge detection without threshold

针对提取图像边缘经常需要设定阈值,而对于光照不均的图像又难以设定合适阈值的问题提出了一种新的边缘检测方法。该方法首先根据对数把图像分解为高频与低频信息,并把对数图像减去其经最大值滤波后的图像提取高频信息,然后根据认知心理学上的Stevens定理,把高频信息转换为心理量。经非最小值抑制细化边缘后,应用Pillar K-means算法提取图像边缘。该方法不需要设定阈值,且对光照不均的图像边缘提取有较好的效果。实验结果证明了该方法的有效性,也表明把图像亮度转换为心理量可以较好地统一不同亮度下的边缘取值。

Concerning the thresholds often being needed in the image edge detection and it is difficult to set good threshold values for the variant illumination image,a new edge detection method was proposed to solve these problems.Firstly,according to the logarithm,an image was decomposed into high frequency and low frequency,and the high frequency image was extracted by the logarithmic image minus the image by the maximum value filter.Then based on the Stevens theorem from cognitive psychology,the high frequency information was transformed into visual psychological quantity.After the edges were thinned by non-minimum suppression,they were extracted by Pillar K-means algorithm.The proposed method has good effect on the variant illumination image and does not need to set threshold value.The experimental results prove the effectiveness of the proposed method,and also show that the edge value in variant intensity may be agreed by converting the intensity to the psychological value.