A collaborative adaptive Wiener filter for image restoration using a spatial-domain multi-patch correlation model

Khaled M Mohamed,Russell C Hardie

doi:10.1186/s13634-014-0189-3

Khaled M Mohamed, Russell C Hardie

Open Access

https://doi.org/10.1186/s13634-014-0189-3

Copy DOI

Abstract

We present a new patch-based image restoration algorithm using an adaptive Wiener filter (AWF) with a novel spatial-domain multi-patch correlation model. The new filter structure is referred to as a collaborative adaptive Wiener filter (CAWF). The CAWF employs a finite size moving window. At each position, the current observation window represents the reference patch. We identify the most similar patches in the image within a given search window about the reference patch. A single-stage weighted sum of all of the pixels in the similar patches is used to estimate the center pixel in the reference patch. The weights are based on a new multi-patch correlation model that takes into account each pixel’s spatial distance to the center of its corresponding patch, as well as the intensity vector distances among the similar patches. One key advantage of the CAWF approach, compared with many other patch-based algorithms, is that it can jointly handle blur and noise. Furthermore, it can also readily treat spatially varying signal and noise statistics. To the best of our knowledge, this is the first multi-patch algorithm to use a single spatial-domain weighted sum of all pixels within multiple similar patches to form its estimate and the first to use a spatial-domain multi-patch correlation model to determine the weights. The experimental results presented show that the proposed method delivers high performance in image restoration in a variety of scenarios.

Highlights

1.1 Image restoration During image acquisition, images are subject to a variety of degradations
We have found that the algorithm is not highly sensitive to these tuning parameters, and good performance can be obtained for a wide range of images using a specified fixed value for these
4 Experimental results we demonstrate the efficacy of the proposed collaborative adaptive Wiener filter (CAWF) algorithm using images with a variety of simulated degradations and using real video frames

Summary

Introduction

1.1 Image restoration During image acquisition, images are subject to a variety of degradations. These invariably include blurring from diffraction and noise from a variety of sources. Restoring such degraded images is a fundamental problem in image processing that has been researched since the earliest days of digital images [1,2]. A wide variety of linear and nonlinear methods have been proposed. A widely used method for image restoration, relevant to the current paper, is the classic Wiener filter [3]. The standard Wiener filter is a linear space-invariant filter designed to minimize mean squared error (MSE)

Methods

Results

Conclusion