Research Of Low Illumination Image Restoration Technology Based On Compressed Sensing

With the rapid development of image processing technology,video surveillance systems have been widely used.In fact,low-illumination images are produced due to insufficient or uneven ambient light.Moreover,these low-illumination images may degrade during acquisition,transmission and conversion because of external factors.So that the amount of available information to people will be further reduced.For the phenomenon of dark images,this paper proposes a Wavelet-Contourlet domain image enhancement algorithm based on improved Retinex.First,the Laplacian filter in Contourlet transform is replaced by wavelet transform,and the image is decomposed by multi-scale.Then,we use a multi-scale Retinex algorithm that introduces a mixed grayscale transformation function to enhance low-frequency images,and use nonlinear functions to enhance high-frequency images.Experiments show that the algorithm can clearly express the detail of the image while enhancing the contrast of the low-illumination image.In order to relieve the image degradation problem,two different restoration algorithms are proposed in this paper:When the point spread function is known,a joint recovery algorithm of compressed sensing and two step iterative threshold shrinkage(TwIST)is proposed in this paper.First,the observation matrix is used to obtain a small amount of data in the high-frequency image.Then,the TwIST algorithm under the compressed sensing theory framework is used to recover high-frequency images,and the GPSR_BB algorithm is used to recover low-frequency images.Experiments show that the proposed algorithm obtains a clearer texture than the traditional algorithm and effectively improves the subjective and objective quality of degraded images at low sampling rates.When the point spread function is unknown,this paper proposes a K-SVD adaptive blind restoration algorithm based on compressed sensing.First,iterative methods are used to alternately estimate and update sparse vectors,sparse dictionaries,and point spread functions.Then,we use the Compressed Sensing and TwIST joint restoration algorithm proposed in this paper to recover the degraded image.The experimental results show that when the algorithm obtains similar result to others,the amount of observation data,storage space,and calculation amount required are less.Low-illumination images not only have low brightness but also have degradation.Therefore,this paper combines the compressed-sensing K-SVD adaptive blind restoration algorithm with the image enhancement algorithm proposed in this paper,and applies it to the restoration of low-illumination images.Experiments show that this method can effectively improve the subjective and objective quality of degraded images under low illumination conditions.