Probabilistic Residual Learning for Aleatoric Uncertainty in Image\n Restoration

2019/08/01 by Chen Zhang, Bangti Jin, Zhang, Chen +1
Decision Sciences · Engineering · Mathematics · #FOS: Computer and information sciences #FOS: Electrical engineering #Image and Video Processing (eess.IV) #Machine Learning (cs.LG) #Machine Learning (stat.ML) #Probabilistic and Robust Engineering Design #Sparse and Compressive Sensing Techniques #Statistical Methods and Inference #electronic engineering #information engineering

paper · pdf · doi:10.48550/arxiv.1908.01010

openalex publication_date 2019/08/01 · openalex created_date 2022/07/28 · openalex updated_date 2026/07/28

Abstract

Aleatoric uncertainty is an intrinsic property of ill-posed inverse and\nimaging problems. Its quantification is vital for assessing the reliability of\nrelevant point estimates. In this paper, we propose an efficient framework for\nquantifying aleatoric uncertainty for deep residual learning and showcase its\nsignificant potential on image restoration. In the framework, we divide the\nconditional probability modeling for the residual variable into a deterministic\nhomo-dimensional level, a stochastic low-dimensional level and a merging level.\nThe low-dimensionality is especially suitable for sparse correlation between\nimage pixels, enables efficient sampling for high dimensional problems and acts\nas a regularizer for the distribution. Preliminary numerical experiments show\nthat the proposed method can give not only state-of-the-art point estimates of\nimage restoration but also useful associated uncertainty information.\n

Citations

• Expectation Propagation for approximate Bayesian inference
• Auto-Encoding Variational Bayes
• A Practical Bayesian Framework for Backpropagation Networks
• Beyond a Gaussian Denoiser: Residual Learning of Deep CNN for Image Denoising
• What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?
• Learned D-AMP: Principled Neural Network based Compressive Image\n Recovery
• Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles
• Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learning

Cited by

• Variational Inference for Computational Imaging Inverse Problems

Related

• Probabilistic Contrastive Learning Recovers the Correct Aleatoric Uncertainty of Ambiguous Inputs
• Principal Uncertainty Quantification with Spatial Correlation for Image Restoration Problems
• Uncertainty Quantification with Generative Models
• Towards Distribution-Shift Uncertainty Estimation for Inverse Problems with Generative Priors
• Uncertainty Quantification in Deep Residual Neural Networks