Extending MC-SURE to denoise sensor data streams

We propose a method to adaptively denoise sensor data streams corrupted by noise that can be approximated as additive white Gaussian. This on-line filtering method is based on the Monte-Carlo Stein´s Unbiased Risk Estimate (MC-SURE) algorithm, which enables a blind optimization of the denoising parameters for a wide class of filters. We first identify the challenges that arise as the MC-SURE algorithm is adapted to on-line data processing. We then propose a framework to address these challenges and demonstrate the application of the algorithm using real-world datasets.