After a decade of extensive study of the sparse representation synthesismodel, we can safely say that this is a mature and stable field, with cleartheoretical foundations, and appealing applications. Alongside this approach,there is an analysis counterpart model, which, despite its similarity to thesynthesis alternative, is markedly different. Surprisingly, the analysis modeldid not get a similar attention, and its understanding today is shallow andpartial. In this paper we take a closer look at the analysis approach, betterdefine it as a generative model for signals, and contrast it with the synthesisone. This work proposes effective pursuit methods that aim to solve inverseproblems regularized with the analysis-model prior, accompanied by apreliminary theoretical study of their performance. We demonstrate theeffectiveness of the analysis model in several experiments.
translated by 谷歌翻译