Stable recovery of sparse signals with coherent tight frames via lp-analysis approach
Dekai Liu,Song Li
DOI: https://doi.org/10.1142/S021953051650010X
IF: 1.9559
2017-01-01
Analysis and Applications
Abstract:In this paper, we consider to recover a signal which is sparse in terms of a tight frame from undersampled measurements via l(p)-minimization problem for 0 < p <= 1. In [Compressed sensing with coherent tight frames via l(p)-minimization for 0 < p <= 1, Inverse Probl. Imaging 8 (2014) 761-777], Li and Lin proved that when 0 < delta(2s) < 1, there exists a (p) over bar (0, 1], depending on delta(2s) such that for any p is an element of(0,(p) over bar], each solution of the l(p)-minimization problem can approximate the true signal well. The constant delta(2s) is referred to as the D-RIP constant of order 2s which was first introduced by Candes et al. in [Compressed sensing with coherent and redundant dictionaries, Appl. Comput. Harmon. Anal. 31 (2011) 59-73]. The main aim of this paper is to give the closed-form expression of (p) over bar We show that for every D-RIP constant delta(2s) is an element of(0, 1), if 0 < p <= <(p)over bar>, where (p) over bar = { min{1, 50/31(1-delta(2s))}, delta(2s) is an element of (0,0.7183), 0.4542, delta(2s) is an element of [0.7183, 0.7729), 2(1-delta(2s)), delta(2s) is an element of [0.7729, 1), then the lp-minimization problem can reconstruct the true signal approximately well. Our main results also hold for the complex case, i. e. the measurement matrix, the tight frame and the signal are all in the complex domain. It should be noted that the D-RIP condition is independent of the coherence of the tight frame (see [Compressed sensing with coherent and redundant dictionaries, Appl. Comput. Harmon. Anal. 31 (2011) 59-73]). In particular, when the tight frame reduces to an identity matrix or an orthonormal matrix, the conclusions in our paper coincide with the results appeared in [Stable recovery of sparse signals via lp-minimization, Appl. Comput. Harmon. Anal. 38 (2015) 161-176].