Tensor methods for low-level vision

Abstract

Low-level vision is a processing system that plays an important role in human as well as in machine visual pattern recognition. It is often used to refer to information processing based on local visual features such as edges, corners, colors, and self-similarity. Typical examples in the research fields of computer vision and image/ signal processing are compression, noise removal, deblurring, superresolution, image inpainting, computed tomography, and compressed sensing. In this chapter, we introduce the tensor representations, mathematical models, and optimization algorithms and illustrate their application to selected low-level vision tasks.