An improved OMP method based on memory effect and its application

In order to improve reconstruction accuracy of the original signal while giving consideration to the computational complexity, an improved Orthogonal Matching Pursuit (OMP)method based on memory effect (MOMP) is proposed. On one hand, in the process of OMP algorithm, only the maximum correlation column is retained when conducting correlation operation, while other columns are neglected and will not contribute to select the final column space. MOMP algorithm deals with this problem by combining correlation operations of each iteration, as a result, the data could be fully utilized. On the other hand, extra addition and multiplication operations are not introduced to the proposed method, the additional part of calculations are simple judgment and comparison operations within each iteration. Thus, by analyzing the complexity and comparing the running time of different algorithms, we can conclude that MOMP is a fast and practical reconstruction method. The experimental results on an ISAR image show that MOMP has a good reconstruction performance and consumes the approximate time of OMP, which has great significances and prospects in the area of compressive sensing.