基于稀疏表示的图像恢复方法

摘要

基于过完备字典的稀疏表示算法能够刻画图像的内在结构和先验属性，实现图像的最优逼近，目前在图像恢复中被广泛应用。但是，为获得精度更高的图像恢复效果，稀疏表示中仍存在一些关键问题需要进一步解决，如字典模型的构建、特征系数的优化稀疏分解、算法的收敛速度、噪声模型的估计、低维数据的线性不可分性等问题。鉴于此，本研究主要讨论基于稀疏表示的图像恢复方法，相关研究结果具有重要的理论研究意义和实用性。%Sparse representation algorithm based on over-complete dictionary can describe images’inner structure and prior properties and realize the optimized approximation so as to be widely used in image processing. However,in order to obtain higher effect of image reconstruction,we still have some key problems to solve in sparse representation,such as the construction of dictionary models,the optimized sparse decomposition,the convergent speed of algorithms,the estimation of noise models,the linearly inseparability of low dimensional data and so on. So,some image restoration methods based on sparse representation are discussed here,and some relevant research results are of theoretical and practical significance.