Research On Edge Detection Algorithms Based On Logarithmic Domain Gradient And Morphology Enhancement

The edge is an important low-level feature information in pattern recognition,and it is an important basis for local feature recognition.The image information obtained by the image signal acquisition device,the amount of information that can be recognized by the image content,is susceptible to shooting angle,occlusion,and illumination intensity during information collection.The problem that image edge information can not be recognized due to the intensity of illumination can be solved by digital image processing technology.Improving the quality of an image with uneven illumination and accurately detecting its edge features will involve the research of image denoising,enhancement and edge detection.In this paper,two different edge detection algorithms are proposed for the problem of unclear image edge recognition under uneven illumination conditions.The specific research contents are as follows:(1)The development history of the first-order operator to the second-order operator of the edge detection algorithm and the research status in the field are briefly introduced.Aiming at the image edge detection problem under the condition of uneven illumination,several commonly used image enhancement algorithms and edge detection algorithms are elaborated,and the advantages and disadvantages of different algorithms are analyzed.Some improvements are proposed in combination with existing detection algorithms.(2)Edge information is very sensitive to illumination,especially when the surface of the object is shaded or the angle of illumination changes,it is not easy to detect the complete edge information,which will cause serious recognition errors.In order to eliminate the influence of uneven illumination on edge detection,an edge detection method combining logarithmic domain gradient and improved Sobel operator is proposed.Firstly,the method attenuates the lowfrequency incident component in the logarithmic domain to enhance the high-frequency reflection component and enhance the image brightness..Then,synthesize the improved 4-direction Sobel operator and express the gradient with infinite norm,making the edges of the image to be detected more complete.Finally,the optimal threshold is determined by linear combination of the Bernsen algorithm threshold and the Gaussian filtering Bernsen algorithm threshold and make the edges of the image more continuous and complete.The simulation results show that the method can effectively eliminate the influence of uneven illumination on image edge detection,and the method is better than other related algorithms and literature in edge detection of different illumination images.(3)In order to improve the enhancement of uneven images,then accurately extract the edge detail features in the image.This paper proposes an edge detection algorithm combining multistructural morphology with maximum entropy threshold.Firstly,the algorithm constructs a structural element with 8 directions and a window size of(2w(10)1)(?2w(10)1),solving the problem that single direction and single scale structural elements are difficult to achieve optimal filtering effect.Then,.the details of bright and dark areas in images are extracted by mathematical morphology operations,and the details of transition areas are further integrated.Finally,combined with the improved maximum entropy threshold segmentation method,the optimal segmentation threshold is obtained to achieve accurate location of edge detail features.The simulation results show that the algorithm has a good detection effect for non-uniform illuminated images,and can achieve better edge detail feature location.