基于改进Canny算法的噪声图像边缘检测

Edge detection of noise image based on improved Canny algorithm

摘要:

针对传统Canny边缘检测算法对噪声图像的去噪效果不佳，以及双阈值需要预先设定的问题，提出了一种基于改进Canny算法的噪声图像的边缘检测。首先构建自适应高斯滤波器对曲度算子进行改进，得到优化的二值边缘图；然后基于最大类间方差法构建了灰度梯度映射函数，确定最佳的双阈值；最后对二值边缘图进行双阈值检测以及边缘连接。实验结果表明，改进算法与现有Canny算法相比，在不同类型噪声和不同浓度噪声的环境下，改进算法提高了对噪声图像边缘检测的性能，其中PSNR值平均提高了1.9%，MSE值平均降低了1.6%，且具有自适应性强、运行效率高的优点。

Abstract:

Aiming at the problem that the traditional Canny edge detection algorithm has poor denoising effect on noisy images and the double threshold needs to be set in advance, a noise image edge detection based on improved Canny algorithm is proposed. Firstly, an adaptive Gaussian filter is constructed to improve the curvature operator to obtain an optimized binary edge map; Then, the gray gradient mapping function is constructed based on the variance between the largest classes, and the optimal double threshold is determined. Finally, double threshold detection and edge connection are performed on the binary edge graph. The experimental results show that compared with Canny algorithm, the improved algorithm improves the performance of noise image edge detection in the environment of different types of noise and different concentrations of noise. The average PSNR value is increased by 1.9%, and the average MSE value is reduced by 1.6%, and the improved algorithm has the advantages of strong adaptability and high operating efficiency.