基于低秩矩阵逼近的图像恢复方法研究-计算机技术专业论文.docxVIP

优秀毕业论文 精品参考文献资料 摘要过对大量的退化高光谱图像序列进行实验，验证了所提方法的有效性，实验结果 摘要 过对大量的退化高光谱图像序列进行实验，验证了所提方法的有效性，实验结果 表明所提算法优于当前流行的高光谱图像恢复方法。 关键词：低秩矩阵逼近；大气扰动去除：高光谱图像恢复； 万方数据 AbstractImage Abstract Image restoration is a basic task in the field of computer vision．It aims at restoring the valuable information from the degraded image．Image restoration is all important technology to make users can see what they want to see and a prerequisite of image analysis and object recognition．Therefore，image restoration is very important for the task of computer vision from the low level to the high level． In the application of image restoration，we focus on the long-distance imaging sensor which makes the images degraded during imaging because of all kinds of degradations．We solve two problems：the atmospheric turbulence removal and the hyperspectral image restoration．The former is challenging because the point spread function changes、航tll the spatial and temporal variation which is difficult to make a model．The degradation of the hyperspectral image(HSI)is of noises，dead line and stripes．It is challenging to restore an image from noises，dead line and stripes simultaneously．Considering the property of the low rank of the image sequence，we implement the low-rank matrix approximation to restore a degraded image．The contributions are as follows： 1．We proposed a novel turbulence removal method．We use robust principal component analysis(RPCA)to generate the original reference image．After that，we iterafively enhance the reference image and registered image frames based on an optimal variation model which combines the non-local similarity prior and the steering kernel regression prior for geometric correction．Then a novel algorithm based on split Bregman algorithm is developed for solving the variation model．Finally,we develop an algorithm based on weight nuclear norm minimization(WNNM)to remove noises of video frames，which is first done for video denoising．The extensive experiments demonstrate that the propose