Research On Compressed Sensing Image Restoration For Transmission Lines

At present,the inspection of transmission lines in China's power grid is mainly completed by unmanned aerial vehicles and manual inspection.Due to the dependence on manual operation and the limitation of hardware equipment itself,it is easily affected by rain,fog and other environmental factors,and the detected images are often not susceptible to Gaussian noise interference.In order to meet the requirements of improving image quality,this paper has found a fast and convenient way to finish the image fuzzy restoration and image denoising after acquiring the low-quality transmission line image,which is all based on sparse representation the image reconstruction theory.Combined with CS theory,this paper conducts a deep research on sparse representation,including the image sparse model research,redundant dictionary algorithm research and application research of image signal sparse representation.To show that the algorithm has significant improvement in image processing,this thesis uses the classical natural images and artificial acquisition image contrast image reconstruction,with the improvement of peak signal-to-noise ratio and image reconstruction image intuitive visual effect of improvement as the goal,key research based on image sparse representation training model and optimization of joint dictionary reconstruction algorithm,and applied it to the actual inspection.1.For image gaussian noise interference,In this method,block grouping weighted coding is combined with sparse non-local regularization,and different weighting processing is established by local sparse and non-local similarity constraints.In order to balance the sparsity of fidelity and regularization terms,adaptive regularization parameters are introduced to establish the denoising model.K-svd algorithm is adopted in dictionary training,and the experimental simulation results clearly show that this algorithm is better than the traditional k-svd algorithm and BM3 D algorithm.This algorithm not only improves the PSNR value,but also maintains better image edges and texture details.The experimental results show that the peak signal noise of the denoising image obtained by this method is 0.7205~2.265 dB higher than the similar method,and the image denoising effect is better.The collected image can also better retain the feature texture information such as details,and achieve a good practical effect.2.Aimed at the shortage problem of corrupt image restoration,using nonlocal similarity,combined with the feature of the structure of natural images,the comparability of nonlocal grouping weighted coding work,improve the algorithm,adaptive fidelity term as a model,and the structure similarity of K-SVD dictionary based on sparse representation in the form of training high and low frequency dictionary optimization,makes the algorithm more has the directivity,further enhance the effect of image restoration.Experimental results show that the proposed PSNR method can improve the image reconstruction quality by 2.58 db on average,which reduces the image reconstruction error,improves the contour and edge of the image,and obtains better image reconstruction quality.