Mixed noise removal in digital images using enhanced TV filters

We propose a technique for constructing adaptively removable filters to filter both Gaussian and impulse noise, based on a Total Variation (TV) filter. This technique offers the benefit of effectively repairing images degraded by the simultaneous superimposition of Gaussian and impulse noise, by referencing positional information about the impulse noise output from an impulse detector at the previous stage and by controlling the parameters of the TV filter. An image restoration experiment conducted by means of a simulation with standard images confirmed that this technique has the potential to deliver significantly better image quality than conventional linear/nonlinear filters.