Edge Detection of Noise Images Based on Improved Canny Algorithm with Adaptive Threshold
Jing Zhang · 2024
Edge detection is a significant research direction in the fields of image processing and computer vision, widely applied in image segmentation, object detection, and other areas. The traditional Canny edge detection algorithm excels in noise resistance and edge localization, but it still faces challenges such as edge information loss and the need for manual setting of dual thresholds when processing noisy images. This paper proposes an improved Canny edge detection algorithm with adaptive thresholds, which effectively addresses the deficiencies of the traditional Canny algorithm in noise suppression and threshold setting by introducing a wavelet filter and an adaptive threshold determination method based on Otsu's method for maximizing the between-class variance. Experimental results on the BSD500 dataset demonstrate that the proposed algorithm performs excellently in edge detection of noisy images, with an accuracy improvement of approximately 5%-10% compared to the traditional Canny algorithm.