Research On Related Issues Of Several Types Of Image Edge Detection

In recent years,with the rapid development of image processing research,the research of non-standard images has gradually become a research hotspot,such as fabric images,remote sensing images and gear images.Compared with standard images,some remote sensing images have complex backgrounds and low light,and noises ware easily mixed in during acquisition;some fabric images have rich texture information,which increases the difficulty of related processing,and it was easy to mix in noise during acquisition.Among the visual perception features,the edge feature was the most basic low-level feature.This feature plays an important role in the simplification and analysis of image information,and lays a good foundation for the deeper processing of subsequent images.And the edge detection after the image was grayed out is our most common image edge detection method,and the color image has a variety of different color spaces,which makes the edge detection more complicated.Therefore,this article will focus on the related problems of image edge detection such as gray-scale printed fabric images,color images,and gray-scale remote sensing images.Aiming at the problem of rich texture information and noise in printed fabric images,this paper proposes an edge extraction algorithm for printed fabric images based on improved filters and wavelets.The idea of coefficient correlation analysis was applied to the improvement of the filter,and the exponential function was combined with the idea of continued fraction approximation to determine the weight.The Euclidean distance formula is used to improve the adaptive median filter,and the adaptive weight formula is used to determine the weight.Finally,the two-dimensional Otsu algorithm is improved to obtain the optimal gradient threshold for the wavelet modulus maximum method.In view of the high correlation between RGB color space components in color image edge detection,some color information cannot be effectively identified,detected,and low noise resistance.This paper proposes a multi-layer wavelet threshold denoising function for color image edges Detection method.The improved wavelet threshold denoising method was used to denoise the color image,and then use the quaternion idea to construct a four-direction feature matrix to solve the gradient amplitude and angle of the Canny operator,and finally perform non-maximum suppression on the gradient amplitude,And adopt adaptive dual-threshold processing to get the edge detection image.Aiming at the problems of low contrast and large amounts of noise in remote sensing images,this paper proposes a remote sensing image edge detection algorithm combined with image enhancement.The basic idea is to use Top-hat transform for image enhancement to highlight the details of the high and low frequency components decomposed by NSCT(non-subsampled contourlet transform),and then use the improved Canny operator to obtain the edge image,and remove some isolated points to obtain the final edge image.Aiming at the problem that it was difficult to effectively suppress noise in the edge detection of noisy gear images,an edge detection algorithm combining Canny operator and mathematical morphology was proposed.Firstly,the image was decomposed by wavelet to obtain the sub-images of each layer,and then adaptive weighted fusion was used for the sub-images,and finally the image was reconstructed to obtain the edge detection image.