Research Of Image Defogging And Image Reconstruction Algorithms

In many application fields such as astronomical imaging,medicine,military,and video coding,image restoration techniques are relied on to overcome image degradation resulting in image degradation,noise interference,image defects,and other problems.Therefore,image restoration techniques have always had a crucial position in image processing and have important theoretical values and significance.In particular,photographs taken in foggy environments usually show visual defects such as color distortion and offset,limited dynamic compression range,and low visual contrast,which cannot show a clear visual appearance and make it difficult to detect and analyze visual targets based on them.In the field of Compressed Sensing(CS),a type of signal sampling that has emerged in recent years,image reconstruction algorithms generally suffer from high computational complexity,low accuracy,and more demanding requirements on the sensing matrix.Therefore,in this paper,two types of algorithms,image defogging and image reconstruction in image restoration,are investigated as follows.(1)An improved fused dark channel defogging method based on multi-scale wavelet transform is proposed for the problems that the dark channel priori algorithm will have color distortion at larger depth of field and is susceptible to noise interference and long running time.Firstly,the fogged image is decomposed into two levels of wavelets,then the high-frequency components are denoised using soft thresholding,and the low-frequency components are defogged using an improved adaptive fusion dark channel.Finally,a local linear model is used to correlate the high and low frequency components coefficients for wavelet reconstruction.The experimental results show that the improved fused dark channel defogging method based on multiscale wavelet transform has high defogging efficiency and can improve the quality of defogged images very well.(2)An adaptive single image defogging method based on sky segmentation is proposed to address the problem that the defogged image obtained from the dark channel priori will have color distortion and be easily disturbed by bright objects.Firstly,the initial atmospheric light value is estimated using the dark and bright channels,the sky region and the exact atmospheric light value are obtained using the information entropy and the dark channel image,and then the initial transmittance is obtained,and then the initial transmittance is estimated accurately using the fast guide filter,and an adaptive weighting factor is introduced to optimize the constraint on the transmittance mapping respectively,and then a bright object error amplification compensation factor is introduced to correct the sky region and the non-sky region respectively.After the atmospheric scattering model to obtain the defogged image,and then a nonlinear mapping method for brightness adjustment,a clearer and more natural defogged image can be finally obtained.The experimental results show that the adaptive single image defogging method based on sky segmentation has better defogging effect for fogged images containing sky areas and bright objects.(3)To address the problems of high computational complexity and low accuracy when reconstructing images using Orthogonal Matching Pursuit(OMP)with discrete cosine transform,we propose an image reconstruction method based on BCS-DMP using the computational complexity advantage of Detouring Matching Pursuit(DMP)and Block Compressed Sensing(BCS).The BCS-DMP image reconstruction method is proposed by taking advantage of the computational complexity of Detouring Matching Pursuit(DMP)and Block Compressed Sensing(BCS).Firstly,the image signal is discrete cosine transformed and uniformly blocked,secondly,the DMP algorithm is used to reconstruct the signal,the discrete cosine inverse transform of the reconstructed signal is used to reconstruct the image,and finally,the mean filtering algorithm is used for smoothing to reduce the block effect of the image.The experimental results show that the peak signal-to-noise ratio of the reconstructed images of the BCS-DMP algorithm is also higher than that of the comparison algorithm when a smaller compression ratio is used,and it has a greater advantage in reconstruction time.The proposed defogging algorithm demonstrates the advantages of adaptive defogging method based on wavelet transform and sky segmentation,which can obtain better quality and better visual effect of defogged images.It also has better defogging effect for fogged images containing sky regions and bright objects.At the same time,compared with other classical defogging methods and deep learning defogging methods,this paper also has certain advantages in several objective evaluation indexes.The proposed BCS-DMP image reconstruction algorithm obtains reconstructed images with higher peak signal-to-noise ratios than those of base tracking and orthogonal matching tracking algorithms,and has higher reconstruction efficiency.