English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Tools annual

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

In recent years, large convolutional neural networks have been widely used astools for image deblurring, because of their ability in restoring images veryprecisely. It is well known that image deblurring is mathematically modeled asan ill-posed inverse problem and its solution is difficult to approximate whennoise affects the data. Really, one limitation of neural networks fordeblurring is their sensitivity to noise and other perturbations, which canlead to instability and produce poor reconstructions. In addition, networks donot necessarily take into account the numerical formulation of the underlyingimaging problem, when trained end-to-end. In this paper, we propose somestrategies to improve stability without losing to much accuracy to deblurimages with deep-learning based methods. First, we suggest a very small neuralarchitecture, which reduces the execution time for training, satisfying a greenAI need, and does not extremely amplify noise in the computed image. Second, weintroduce a unified framework where a pre-processing step balances the lack ofstability of the following, neural network-based, step. Two differentpre-processors are presented: the former implements a strong parameter-freedenoiser, and the latter is a variational model-based regularized formulationof the latent imaging problem. This framework is also formally characterized bymathematical analysis. Numerical experiments are performed to verify theaccuracy and stability of the proposed approaches for image deblurring whenunknown or not-quantified noise is present; the results confirm that theyimprove the network stability with respect to noise. In particular, themodel-based framework represents the most reliable trade-off between visualprecision and robustness.

Ambiguity in solving imaging inverse problems with deep learning based operators