Adaptively determining regularisation parameters in non-local total variation regularisation for image denoising

A non-local selection scheme of regularisation parameters is exploited for non-local total variation (NLTV) regularisation, based on a content-aware function that can determine regularisation parameters through discriminating image contents, with the aid of a non-local structure tensor calculated using the weights from the original NLTV.

Experimental results on simulated and real noisy images show the relatively good performance of the proposed scheme based on NLTV, in terms of visual improvement and peak signal-to-noise ratio value.

• Share: