Research On Atmospheric Blur Removing Algorithms Of Remote Sensing Images

At present, imaging the interested object through atmosphere is the inevitable problem for all the imaging equipments running in the Earth's atmosphere. For the existence of atmospheric structure such as turbulence, the images obtained by these equipments are often blurred in varying degrees. Consequently, these atmospheric blurred images require to be restored. The atmospheric blurred degraded image restoration and reconstruction is an interdisciplinary, cross-cutting issue as well as a global problem which the researchers in related fields at home and abroad are trying to solve. The difficulty to restore degraded images caused by atmospheric disturbances lies in its unknown and varying randomly degradation model, which can hardly be expressed by precise mathematical formulas or models. What's worse, noise contained in the blurred images further adds the difficulty to restore such images. The traditional image restoration algorithms are pointed to the circumstances that the degradation models are knew, whereas the study of image restoration algorithms with degradation unknown is a challenging research field, which owns a broad application market and prospects. Especially with the development of space exploration technology and the advancement of space resolution of earth observation system, the atmospheric interference are becoming a main constraint to further improve the resolution of remote sensors, therefore, to solve the problem is increasingly needed.Above all, based on the summarization of the existing algorithms home and abroad about atmospheric turbulence effect image restoration, with the support of the National High Technology Research Program (863 Program, No.: 2006AA12Z110), this paper studied and discussed the high-resolution restoration and reconstruction of remote sensing images affected by atmospheric blur. Via modifying the existing turbulence elimination algorithms and combining the turbulence restoration technology commonly used in astronomical observations with remote sensing images, the paper brought up a novel way to estimate atmospheric MTF according to localized weather data at imaging moment, and verified the idea with remote sensing data observed by domestic and foreign remote sensing satellites. The research results and contributions of the paper represent mainly in:1) The design of estimation algorithms of turbulence MTF, aerosol MTF and atmospheric MTF based on meteorological data. The atmospheric overall modulation transfer function MTF is mainly composed by the product of atmospheric turbulence MTF and aerosol MTF, while the atmospheric MTF and aerosol MTF meteorological parameters can be described by meteorological parameters (Refractive index structure coefficient) and the aerosol size distribution respectively. Recorded under imaging time regional, using to the turbulence MTF, aerosol MTF and atmospheric overall MTF are estimated by these two parameters, according to the weather data of the same region at imaging moment;2) Imaging data MTF estimation based on image data. Combined with the physical meaning of MTF the two current accesses to the optical imaging system MTF are introduced and experimented. The restoration of images observed by optical imaging system and evaluation to performance of optical imaging system are realized by use of MTF decline in frequency domain.3) Improved Wiener filter algorithm combined with estimated atmospheric MTF. The Wiener filter are modified with the use of estimated atmospheric MTF, and become more suitable for remote sensing image restoration.4) Improved direct deconvolution algorithm combined with estimated atmospheric MTF. The direct deconvolution algorithms are modified with the use of estimated atmospheric MTF, and become more suitable for remote sensing image restoration.5) Adaptive remote sensing image restoration algorithm based on "myopic" deconvolution algorithm. The "myopic" deconvolution which is widely used in microscopic observation and astronomical image processing is introduced to restoration and reconstruction of atmospheric remote sensing images. Based on the realization of the original "myopic" deconvolution algorithm, an adaptive image deconvolution algorithm for remote sensing is proposed according to its own characteristics. The algorithm applies the conjugate gradient optimization methods, including a balance of maximum likelihood estimates and the adaptive object regularization scheme, so the algorithm obtained satisfactory results while the run time and efficiency has also been significantly improved.6) Deconvolution algorithm of remote sensing images based on Bayesian principle. The unknown scene are modeled by key variable model and the scale-invariant random process determined by the global energy scale, and then a Bayesian approach to estimate the blur and noise parameters in the models based on probability and statistics theory is given. According to the estimated parameter values and the degradation model, the blurred degradation function MTF of imaging system can be obtained, and images are restored by the traditional restoration methods.7) Gradient domain high dynamic compression based remote sensing image enhancement algorithms. Pointed to the display problems of high radiometric resolution remote sensing images on general display device, a new algorithm of displaying high radiometric resolution remote sensing images is given. The results show that the algorithm manages to realize high dynamic range compression as well as retain image details and suppress common edge effect. The algorithm owns a simple theory, computational efficiency and ease of use. The effectiveness of the algorithm is verified by processing the real high radiation resolution remote sensing image with dynamic compression.