Font Size: a A A

Research On Image Quality Improvement And Assessment For Optical Remote Sensing Image

Posted on:2017-05-24Degree:DoctorType:Dissertation
Country:ChinaCandidate:G M CuiFull Text:PDF
GTID:1108330491962875Subject:Optical Engineering
Abstract/Summary:PDF Full Text Request
Optical remote sensing imaging is an important technology of information detection. It relies on received radiation of the target itself to realize information acquisition, which obtains a significant application value in military and civil fields, such as urban planning, environment monitoring, resource exploration, air defense early warning and target recognition. In the process of remote sensing image chain, the remote sensing images are easy to be affected by various degradation factors, such as target radiation, atmospheric disturbance, imaging system, optical-to-electric signal transformation and vibration of satellite platform. These factors will introduce different degrees of degradation into remote remote sensing images, led to a decline in image quality. This would affect the subsequent image processing and seriously limit the application of remote sensing images. The means of on-orbit equipment optimization is extremely costly with long period and limited effects. Under the condition of existing hardware, how to utilize the software processing to propose effective methods to improve the quality of remote sensing images has been the focus of experts’attention. Besides, to establish objective perfect remote sensing image quality evaluation system also has the important meaning, which helps provide a guidance for the adjustment of the parameters of satellite and improvement algorithm. In this paper, the degradation factors in the remote sensing chain are analyzed. Combined with the visual features, researches have been focued on the noise estimation for remote sensing images and noisy image quality assessment. Several image quality improvement processing algorithms are proposed. Moreover, aimed at the task of moon remote sensing image quality improvement, specific image restoration solutions and evaluation methods are designed to meet the target demand.The optical remote sensing imaging chain model is analyzed, including ground objects model, atmospheric radiation transfer model, camera model and the satellite orbit attitude model. We have discussed the major degradation factors for optical remote sensing imaging and analyzed the causes of image degradation to each kind of degradation factor. The corresponding mathematical model is established.We have carried out deep researches on remote sensing image quality improvement. A modified Richardson-Lucy (RL) deblurring method for single image with adaptive reference maps is proposed. By introducing the two steps of adaptive reference maps estimation, the image edge information is calculated more precisely. The algorithm obtains a good performance on ringing suppression and detail preservation. We have proposed a high quality image-pair-based deblurring approach by utilizing the blurred image with long exposure time and noisy image with short exposure time of the same target scences. The local constraint prior and saliency weighted map are introducd to effectively enhance the image details in the restored result as well as suppress the nosie and ringing artifacts. To solve remote sensing image denoising problem, we have presented a remote sensing image denoising method based on non-subsampled Contourlet transform and relative total variation. Also, in order to well utilize the image information from different remote sensing bands, we have designed an image information improvement method based on multi band remote sensing image fusion. By using regional saliency extraction and multi-scale image decomposition, we carried out multi band image fusion. The fusion result retains and enhances the information of different band images effectively and significantly improve image information.The typical image quality evaluation methods are summarized. We have also introduced some kinds of specific evalution parameter for remote sensing images and analyzed the characteristics and available field of several remote sensing image signal-to-noise ratio (SNR) calculation methods. In order to better realize the effective separation of image signal and noise components, the image signal affine reconstruction model is built. The nosiy image is segmented into several image patches with similar size and then the affine reconstruction model is solved to obtain the image signal. Based on this technology, an image noise and SNR estimation method based on noise level accumulation is proposed. The distribution of intensity-noise samples is calculated and the noise level accumulation value and SNR accumulation value are obtained by the weighted summation of noise standard deviation in each image intensity interval. Combined with the human visual contrast sensitive function and affine reconstruction model, an image quality assessment method for noisy images is presented. The proposed metric has a superior performance of the subjective and objective consistency and accuracy on all of the image data bases.We have carried out researches on the image restoration of moon remote sensing image with large amount of image motion blur. We have analyzed the mission requirements of moon surface imaging and image characteristics and worked out the whole experiment scheme and two kinds of specific image restoration approaches. Theoretical simulation experiment and scene shooting simulation experiment are done in order to verify the effectiveness of the restoration algorithms. For the real moon remote sensing blurred images with different levels of image motion, we have used the deblurring methods to restore the images. We have analyzed the compression block effect of the original blurred image and show that the restoration methed will remove the block effect. Besides, a no-reference comprehensive image quality assessment method is designed to evaluate the processed results of the real moon images. Experiments have demonstrated that the restoration algorithms can significantly improve image quality...
Keywords/Search Tags:optical remote sensing image, remote sensing imaging chain, image quality improvement, image restoration, image fusion, noise estimation, signal-to-noise ratio calculation, image quality assessment
PDF Full Text Request
Related items