Fast L0-Based Sparse Signal Recovery
Event details
| Date | 17.01.2011 |
| Hour | 13:30 |
| Speaker | Prof. Nick Kingsbury, Signal Processing and Communications Research Group, University of Cambridge |
| Location | |
| Category | Conferences - Seminars |
We develop an algorithm for reconstructing signals from limited observations in a linear system. We assume an adaptive Gaussian model for the signals, which are assumed to be sparse in some suitable domain (e.g. the complex wavelet domain). This model results in a least squares problem with an iteratively reweighted L2 penalty that is shown to approximate the L0-norm. We propose a fast algorithm to solve the problem within a continuation framework, in order to encourage convergence to the globally optimum solution. In our examples, we show that the correct sparsity map and sparsity level are gradually learnt during the iterations even when the number of observations is reduced, or when observation noise is present. In addition, with the help of interscale signal models, the algorithm is able to recover signals to a better accuracy and with reduced numbers of observations, when compared with L1 and reweighted-L1 methods. The algorithm is applicable to a wide range of input data, including high-resolution 2D images or 3D datasets and can be used to achieve various forms of super-resolution.
Practical information
- General public
- Free