Adaptive Image Denoising by Mixture Adaptation

TL;DR

This paper introduces an EM-based adaptive learning algorithm for patch-based image priors that improves denoising performance by tailoring a generic prior to specific noisy images, outperforming existing methods.

Contribution

It provides a rigorous derivation of the EM adaptation algorithm for image denoising and demonstrates its effectiveness and computational efficiency improvements.

Findings

The EM adaptation algorithm improves denoising results over non-adaptive methods.

It outperforms several state-of-the-art denoising algorithms.

Pre-filtering enhances EM adaptation when the clean image is unavailable.

Abstract

We propose an adaptive learning procedure to learn patch-based image priors for image denoising. The new algorithm, called the Expectation-Maximization (EM) adaptation, takes a generic prior learned from a generic external database and adapts it to the noisy image to generate a specific prior. Different from existing methods that combine internal and external statistics in ad-hoc ways, the proposed algorithm is rigorously derived from a Bayesian hyper-prior perspective. There are two contributions of this paper: First, we provide full derivation of the EM adaptation algorithm and demonstrate methods to improve the computational complexity. Second, in the absence of the latent clean image, we show how EM adaptation can be modified based on pre-filtering. Experimental results show that the proposed adaptation algorithm yields consistently better denoising results than the one without adaptation and is superior to several state-of-the-art algorithms.