【摘要】 图像稀疏分解是图像压缩处理的一种有效的重要环节,但该方法在实现上存在原子匹配收敛与运算耗时的难题。针对基于匹配追踪的稀疏分解,采用一种新的各向异性原子来构造冗余字典,并利用原子能量中间支撑部分来估计原子与待处理图像的内积,从而实现图像的快速稀疏分解。图像分解与重建的结果表明,提出的方法在重建图像质量基本不变的情况下,具有图像稀疏分解表达好,稀疏分解的速度提高1倍。更多还原

【Abstract】 Image sparse decomposition is an efficient and key step for image compression process,but the atom matching convergence and time-costing in computation are its poser in realization.In the paper the author uses a new anisotropy atom to construct over-complete dictionary in light of the sparse decomposition based on matching pursuit,and makes use of an atom energy support to estimate the inner product of atom and image to be treated,thus the fast sparse decomposition of image realized.The results of image decomposition and reconstruction show that the new method expresses image’s sparse decomposition well when keeping the quality of reconstructed image basically unchanged and doubles the sparse decomposing speed.更多还原