Abstract:Denoising stationary process $(X_i)_{i \in Z}$ corrupted by additive white Gaussian noise is a classic and fundamental problem in information theory and statistical signal processing. Despite considerable progress in designing efficient denoising algorithms, for general analog sources, theoretically-founded computationally-efficient methods are yet to be found. For instance in denoising $X^n$ corrupted by noise $Z^n$ as $Y^n=X^n+Z^n$, given the full distribution of $X^n$, a minimum mean square error (MMSE) denoiser needs to compute $E[X^n|Y^n]$. However, for general sources, computing $E[X^n|Y^n]$ is computationally very challenging, if not infeasible. In this paper, starting by a Bayesian setup, where the source distribution is fully known, a novel denoising method, namely, quantized maximum a posteriori (Q-MAP) denoiser, is proposed and its asymptotic performance in the high signal to noise ratio regime is analyzed. Both for memoryless sources, and for structured first-order Markov sources, it is shown that, asymptotically, as $\sigma$ converges to zero, ${1\over \sigma^2}E[(X_i-\hat{X}^{\rm Q-MAP}_i)^2]$ achieved by Q-MAP denoiser converges to the information dimension of the source. For the studied memoryless sources, this limit is known to be optimal. A key advantage of the Q-MAP denoiser is that, unlike an MMSE denoiser, it highlights the key properties of the source distribution that are to be used in its denoising. This property dramatically reduces the computational complexity of approximating the solution of the Q-MAP denoiser. Additionally, it naturally leads to a learning-based denoiser. Using ImageNet database for training, initial simulation results exploring the performance of such a learning-based denoiser in image denoising are presented.

Discrete denoising for channels with memory

Frequency-Relevant Residual Learning for Multi-Modal Image Denoising.

Joint Channel Estimation and Data Recovery of Communication Systems with Sub-Nyquist Receiver.

Denoising of structured random processes

An Induced Additive-Noise Model for Non-Coherent Discrete-Time Memoryless Rayleigh Fading Channels

An Induced Additive-Noise Model for Memoryless Rayleigh-fading Channels

Denoising as well as the best of any two denoisers

Non-Markovian Noise Suppression Simplified through Channel Representation

Neural Network Based Denoising in the Wireless Channel Characterization

A Novel Denoising Method Based on Machine Learning in Channel Measurements

Joint Channel Estimation and LDPC Decoding over Time-Varying Impulsive Noise Channels.

Denoising Noisy Neural Networks: A Bayesian Approach with Compensation.

Channel Estimation via Successive Denoising in MIMO OFDM Systems: A Reinforcement Learning Approach

Information Rates of Channels with Additive White Cauchy Noise

Strong Data Processing Inequalities for Input Constrained Additive Noise Channels

Deep Residual Learning for OTFS Channel Estimation with Arbitrary Noise

Achieving near-Capacity on Large Discrete Memoryless Channels

Efficient Residual Shrinkage CNN Denoiser Design for Intelligent Signal Processing: Modulation Recognition, Detection, and Decoding

Conditional Denoising Diffusion Probabilistic Models for Data Reconstruction Enhancement in Wireless Communications

Dephasing superchannels

Joint Design of Denoising Diffusion Probabilistic Models and LDPC Decoding for Wireless Communications