On B-Spline Framelets Derived from the Unitary Extension Principle

Abstract

The spline wavelet tight frames considered in [A. Ron and Z. Shen, J. Funct. Anal., 148 (1997), pp. 408-447] have been used widely in frame based image analysis and restorations (see, e.g., survey articles [B. Dong and Z. Shen, MRA-based wavelet frames and applications, IAS Lecture , there are few other properties of this family of tight frames that are currently known. The aim of this paper is to present a few new properties of this family that will provide some reasons why it is efficient in image analysis and restorations. In particular, we first present a recurrence formula for computing the generators of higher order spline wavelet tight frames from lower order ones. This simplifies the computations of the exact values of the functions in applications. Second, we represent each generator of spline wavelet tight frames as a certain order of derivative of some univariate box spline that satisfies a few additional properties. This is a crucial property used in [J.-F. Cai et al., J. Amer. Math. Soc., 25 (2012), pp. 1033-1089, where connections between total variational and wavelet frame based approaches for image restorations are established. Finally, we further show that each generator of sufficiently high order spline wavelet tight frames is close to a derivative of a properly scaled Gaussian function. This leads to the result that the wavelet system generated by finitely many consecutive derivatives of a properly scaled Gaussian function forms a frame whose frame bounds can be almost tight.

Image restoration: Total variation, wavelet frames, and beyond

“…In Section 4.3, we will provide details on the quantities defined above using Haar and piecewise linear framelets as examples. The existence of ϕ i is guaranteed by the work [56]. In fact, a uniform form of such a function ϕ i for an arbitrary B-spline framelet is given by [56].…”

“…The existence of ϕ i is guaranteed by the work [56]. In fact, a uniform form of such a function ϕ i for an arbitrary B-spline framelet is given by [56]. (3) When multi-level framelet decomposition is used, i.e.…”

An edge driven wavelet frame model for image restoration

“…where * denotes the discrete convolution. The key observation made by [11] is that for the piecewise Bspline framelets ψ α , there exists a function ϕ α associated to ψ α such that R 2 ϕ α dx = 0, ψ α = ∂ α ϕ α and supp(ψ α ) = supp(ϕ α ), and the explicit formulae of ϕ α are given in [57]. With the aid of the theory of distribution [42,55], Proposition 2.1 generalizes the same result to any tensor product framelet.…”

Framelet pooling aided deep learning network: the method to process high dimensional medical data

“…Interested readers may consult e.g. (Ron and Shen 1997, Dong and Shen 2012, Shen and Xu 2013 for the detailed surveys. In this paper, we only consider the 2-dimensional case as we focus on the undersampled 2-dimensional image reconstruction.…”