Optical Remote Sensing Motion Degration Simulation And Image Restoration Technology

Optical remote sensing relies on receiving the target and the radiation information of environmental light to obtain information.It covers a wide range,and can be used to enrich the spectrum,which plays an important role in military,civil and commercial fields.The vibration of the satellite has become a very important factor affecting the quality of satellite imaging with the decrease of the height of the spacecraft orbit,the increase of space camera's focal length,the increase of the ground resolution and the smaller pixel size of the CCD.The micro vibration of the satellite will cause optical pointing jitter while the space remote sensing camera is in the imaging process,thus resulting in the degration of image quality,image blur and geometric distortion.In the paper,based on the proposed model of the optical remote sensing imaging motion degradation simulation,we focus on improving the quality of blurred remote sensing images by image restoration technology,and some key problems related are discussed in this paper in detail as follows.Based on the analysis of the imaging mechanism of different types' of remote sensing cameras,we have carried out the motion imaging simulation of optical remote sensing satellite.Through the establishment of an overall camera and photographic condition model and realted relationship of the vibration on the image plane,the optical satellite imaging simulation for satellite micro vibration software is completed.The software can be used to simulate and evaluate the remote sensing degraded images under the condition of satellite micro vibration.We also established a simulation model of random vibration.From the aspect of power spectrum estimation,the theoretical expression formula of the actual random vibration curve is deduced from the power spectral density curve.The simulation results show that the image quality can be improved by suppressing the vibration frequency.We carried out the research of remote sensing image restoration technology.Measurement precision and sampling rate of the motion trajectory during the recovery of remote sensing image are studied and allowable measurement error and sampling rate are given.For remote sensing images which are difficult to obtain an accurate kernel by measurement,a blind restoration method based on improved LO norm is proposed.In the kernel stage,an approximate LO norm is used as the image restoration constraint,which can preserve salient edges and filter the small details of the image.After derivation,LO norm can be approximately converted to a piecewise function that can be directly solved and L1 norm term which can be solved by the Split-Bregman method.Experimental results show that the proposed method is able to estimate relatively complex blur kernel and can obtain good restoration results.We also presents a fast recovery method of TDI scan image.The algorithm employs the highly coincidence of the trajectory of the corresponding adjacent rows in the image,the point spread function(PSF)of each row can be quickly estimated.By using the characteristic that the adjacent row of blurred image corresponds to the slow change of blurred kernel,the restoration of the blurred image row by row can be converted to the image block restoration.The restoration time of the TDI blurred image using the proposed method is significantly reduced.Some special cases during image restoration which do not conform to the concolutional model in the process of deblurring are studied.We have studied the problem of image restoration with partially saturated pixels in the blurred image,and proposed a method to deal with the negative effects of saturation pixel during restoration under the modified multi-frame blind deconvolution framework and a light streak detection scheme is incorporated into the regularization constraint.We also studied the problem that the restoration result of compressed blurred image will enlarge the blocking artifacts and proposed a compressed degraded image blind deconvolution algorithm based on block effect suppression.The method improves the kernel estimation by adding a kernel gradient norm constraint,and in the final restoration stage,a new L2 norm regularization term was added on the basis of the total variation constraint on the image which can suppress the block effect step by step in the iterative process.In the end,this paper studies the image restoration technology of large image motion blur.The key problem is to solve the problem of the large image motion blur due to the high velocity of the spacecraft in the lower orbit while.capturing the lunar surface during the exposure time.According to the characteristics of lunar images,a kernel estimation method based on image morphology is proposed under a high rate of blurred lunar image compression.Also,a restoration method under the Poisson noise model and filters trained by the Markov Radom Field is proposed considing the cosmic background radiation.The massive restoration processing of real captured images of the lunar surface is carried out,and a significant improvement in image quality is achieved.