Univariate Total Variation Denoising

Total variation denoising can be used to approximate a given sequence of noisy observations by a piecewise constant sequence, with adaptively-chosen break points. An efficient linear-time algorithm for total variation denoising is provided here, based on Johnson (2013) .

tvdenoising

The tvdenoising package provides an implementation of an efficient linear-time dynamic programming algorithm for univariate total variation denoising (also called fused lasso signal approximation), due to Johnson (2013), which computes the exact solution, for a given regularization level $\lambda$. You can also find a concise explanation of the algorithm at this link.

Installation

To install the released version from CRAN:

install.packages("tvdenoising")

To install the development version from GitHub:

# install.packages("pak")
pak::pak("glmgen/tvdenoising")