Improved Sobel Algorithm Image Edge Detection Based on Adaptive Threshold and Optimised Convolutional Kernel

The traditional Sobel algorithm has some problems in image edge detection, such as sensitivity to noise and poor adaptability to fixed thresholds. In this paper, an improved Sobel algorithm is proposed, which enhances the edge feature extraction ability by designing an optimised convolutional kernel, and adopts adaptive threshold processing to dynamically determine the threshold according to the local characteristics of the image. Experimental results show that the improved algorithm has significant improvement in noise immunity and edge positioning accuracy, and can be effectively applied to edge detection of complex images.