On Choosing Wavelet Resolution in Image Deblurring

In wavelet deblurring of noisy images the finest resolution levels are key parameters which require fine tuning. In this paper we present a blockwise thresholding rule in the Fourier domain to derive near-optimal finest resolution levels for a large class of blurred-image scenarios. Numerical results based on the WaveD fast algorithm of [5] illustrate the improvement over the current WaveD setting.