The Application Of Coarse And Fine Two-stage Denoising Algorithm Based On Noise Localization In Infrared Image

As an emerging technology,infrared imaging technology is used in many fields such as circuit detection or night target recognition.Most device failures in power systems are accompanied by heating,the infrared image can intuitively reflect the temperature or temperature change of the current scene,therefore,infrared imaging technology can accurately locate the faulty part of the power system.Infrared imaging technology is applied to the circuit detection part of the power system to achieve fault prediction and higher accuracy fault detection,which has guaranteed the smooth progress of production and life.However,infrared images are more susceptible to noise interference than other images due to their small coverage and low contrast.At present,most denoising algorithms cannot judge whether a pixel is noise or not,but all pixels are processed with the same rules.When this kind of denoising method deals with images with more details or the small difference between the subject and the background,it will produce too much filtering of the original image information or can not effectively filter the noise.In order to improve the above mentioned defects,a two-stage image denoising algorithm is proposed and applied to the infrared image of the power system.First,edge refinement and adaptive thresholding are used as preprocessing for denoising.The edge information of noisy images is obtained,the edge information is saved and the edges of the original noise image are deleted to obtain an edgeless noise images.Thresholds₁^?and₂^?are obtained according to the gray value distribution of the image.The image is divided into three regions,and mean values of gray values in different regions are obtained at the same time.Then the three gray-scale average values are taken in order to obtain adaptive thresholds₁ and₂;Secondly,the singular value matrix is obtained by singular value decomposition without edge noise images.The percentage threshold valueis obtained by the characteristics of the singular value matrix itself,and the matrix is deranked by the threshold.The image with rough noise filtering is composed by inverse singular value decomposition of singular value matrix and original left right singular value matrix;Thirdly,according to the gray value distribution of the image,image is divided into"Dark Area","Light Area"and"Gray Area"by adaptive thresholds₁ and₂.A superpixel-like algorithm is proposed by analyzing the idea of superpixel algorithm,it is applied from image segmentation to image denoising,fine noise filtering is completed by filtering out the noise in"Dark Area","Light Area"and"Gray Area";Finally,the extracted image edge information is fused with the denoised image to obtain a denoised image.By comparing the peak signal-to-noise ratio?PSNR?of various algorithm performance indicators,it is verified that the improved algorithm has a good denoising effect;The improved algorithm is applied to the infrared image of the power system.By comparing the PSNR of various algorithms,it is verified that the algorithm effectively removes the noise while retaining the detailed information of the image to a great extent.