The main target of this paper is to research of dictionary learning algorithm used in sparserepresentation.Along with the rapid development of compressive sensing, the sparse representation ofsignals has become a hot research field. The overcomplete dictionary performs good in repre-sentation of signals, and is easy to represent a complex signal in a sparse coding. The overcom-plete dictionary designed by learning algorithm can represent a certain kind of signal adaptivelyto get a more sparse representation. This sort of problems have a wide application and highresearch value.This article conduct a study on designing a learned dictionary algorithm. In the research,traditional dictionary learning algorithm wasted a great quantity of computational complexity inthe early sparse representation stage, and may not achieve optimal for the overcoding of repre-sentation stage.Aiming at these problems, firstly proposed Global Support Alternative Pursuit K-SVDalgorithm(GSAP-KSVD), and improved the algorithm to obtain Alternative Pursuit Guided K-SVD(APG-KSVD) algorithm. In the end, this article experiment the proposed algorithm inimage denoising, and verified the advantage of proposed algorithm. |