Double Sparseness And Its Application In Edge Detection Of Complex Images

People are surrounded by various signals,and image signals are a very common type of signal.With the realization of computer visualization and the development of digital media technology,the research on digital image processing has attracted more and more attention from scholars.Many branches have emerged in the development of image processing,such as image compression,image restoration,image segmentation,and image recognition.In image segmentation,image recognition and other processing methods for feature extraction,edge detection technology is needed to facilitate data reduction and information purification.In the process of edge detection,how to eliminate the influence of non-coherent factors has always been the focus and difficulty of research.The existing edge detection algorithms have limited robustness,so the credibility and accuracy of the detection results depend on the conciseness of the image morphology.This dissertation uses the feature extraction function of the sparse decomposition method to remove irrelevant information,and then detects the edges to improve the accuracy of the edge detection results.Aiming at the problem of large amount of image information and difficult to find regularity in the time domain and frequency domain,this study uses a double sparse dictionary learning scheme to solve the adaptive dictionary corresponding to the image,and performs sparse and redundant representation of the image on the adaptive dictionary.The dual sparse dictionary learning scheme uses wavelet transform to pre-decompose the dictionary learning samples to obtain the high frequency part of the sample,and then perform K-Singular Value Decomposition dictionary learning to obtain the learning dictionary of the image,and perform sparse representation of the original image.The improved dictionary learning method reduces the calculation amount of dictionary learning and improves the efficiency of K-Singular Value Decomposition dictionary learning.In view of the overlapping of texture information in complex images and serious interference to edge detection,this study proposes an edge detection method combined with double sparseness.The sparse and redundant representation methods are adaptive.The feature information in the image can be decomposed and the corresponding sparse coefficients can be obtained,and then the sparse coefficients of the low frequency part can be separated according to the activity of the dictionary atoms,and the purer image of the low-frequency can be reconstructed,and then detect the edge of the image.This method reduces the amount of calculation for edge detection,reduces the complexity of image morphology,and obtains more complete edges.In addition,this study has improved the dictionary atom selection method of the Orthogonal Matching Pursuit algorithm,and the efficiency of the improved algorithm has been improved to a certain extent.