基于原子能量特性的快速图像匹配追踪

The computational load in Matching Pursuit (MP) of images is very huge, and this is the main drawback with MP of images. In this paper, a new fast algorithm is presented to speed up MP of images. Energy distribution of atoms used in MP is studied and most of computational loads can therefore be avoided by making use of the atom energy property. Experimental results show that the MP algorithm developed according to this property is many times faster than the conventional MP method and at the same time the quality of reconstructed images by the new algorithm is as good as by the conventional MP method.】稀疏分解在图像处理中应用的关键障碍之一是图像稀疏分解速度十分缓慢。针对这一问题，提出了一种新的图像稀疏分解的匹配追踪快速算法。研究了图像稀疏分解中使用的原子的能量分布特性，根据原子能量的分布特性，图像匹配追踪中的绝大部分的计算可以省略，因而极大地提高了图像匹配追踪的计算速度。实验结果表明，新的算法比传统的图像匹配追踪算法速度提高了许多倍，而恢复图像的质量没有任何的降低。