2020/01/12 by Amir Averbuch, Averbuch, Amir, Pekka Neittaanmäki +7
Computer Science · Engineering · #Advanced Numerical Analysis Techniques #FOS: Electrical engineering #Image and Signal Denoising Methods #Image and Video Processing (eess.IV) #Medical Image Segmentation Techniques #electronic engineering #information engineering
paper · pdf · doi:10.48550/arxiv.2001.04899
openalex publication_date 2020/01/12 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
The paper presents a new algorithm for the image inpainting problem. The\nalgorithm is using a recently designed versatile library of quasi-analytic\ncomplex-valued wavelet packets (qWPs) which originate from polynomial splines\nof arbitrary orders. Tensor products of 1D qWPs provide a diversity of 2D qWPs\noriented in multiple directions. For example, a set of the fourth-level qWPs\ncomprises 62 different directions. The properties of the presented qWPs such as\nrefined frequency resolution, directionality of waveforms with unlimited number\nof orientations, (anti-)symmetry of waveforms and windowed oscillating\nstructure of waveforms with a variety of frequencies, make them efficient in\nimage processing applications, in particular, in dealing with the inpainting\nproblem addressed in the paper. The obtained results for this problem are quite\ncompetitive with the best state-of-the-art algorithms. The inpainting is\nimplemented by an iterative scheme, which, in essence, is the Split Bregman\nIteration (SBI) procedure supplied with an adaptive variable soft thresholding\nbased on the Bivariate Shrinkage algorithm. In the inpainting experiments,\nperformance comparison between the qWP-based methods and the state-of-the-art\nalgorithms is presented.\n