Total variation wavelet thresholding (Q2385515): Difference between revisions

The authors consider the noise removal and reducing edge artifacts generated by wavelet thresholdings in image denoising and compression. It is known that wavelet thresholdings may generate oscillations near discontinuities. Since about 1990, partial differential equations (PDE) models have been used in image processing in the pixel domain [see, e.g., \textit{T. F. Chan} and \textit{J. Shen}, Image Processing and Analysis. Variational, PDE, wavelet and stochastic methods. SIAM, Philadelphia (2005; Zbl 1095.68127)]. The authors use PDE techniques, specifically total variation minimization, to reduce the oscillations in wavelet thresholdings to select and modify the retained wavelet coefficients so that the reconstructed images have fewer oscillations near edges while the noise is smoothed. Numerical tests with 1-D and 2-D data show that this approach improves the reconstructed signal/image in wavelet compression and in denoising.