Occam filters for stochastic sources with application to digital images

Balas K. Natarajan, Konstantinos Konstantinides, Cormac Herley

Abstract: An Occam filter employs lossy data compression to separate signal from noise. Previously, it was shown that Occam filters can filter random noise from deterministic signals. Here, we show that Occam filters can also separate two stochastic sources, depending on their relative compressibility. We also compare the performance of Occam filters and wavelet-based denoising on digital images.