Contents 1 Introduction
Edge refers to the place where the local brightness of an image changes most significantly, and it exists between two or more regions with different gray values. The current image edge detection methods can be divided into three types [1]-[3]: the first is based on spatial domain differential convolution or some operators similar to differential convolution, such as Sobel operator, Canny operator and Roberts operator; the second is an image segmentation algorithm based on energy minimization based on the energy perspective, including neural network analysis and relaxation algorithms; the third is a segmentation algorithm represented by high-tech that has only been developed in recent years, such as wavelet transform. Because the edge information of the face image is complex under certain conditions and the existence of noise is required, a new test of the adaptability, stability and effectiveness of the edge detection algorithm is proposed. Although some new edge detection methods have been proposed, due to the advantages of fast computing speed and universality, the classical space-based convolution detection operator has still been widely used in edge detection of face images.
