Image Fusion

Algorithms and Applications

Tania Stathaki

• Publisher: Academic Press
• Year: 2011
• Print ISBN: 9780123725295
• eBook ISBN: 9780080558523
• Topic: Technology & Engineering
• Subtopic: Digital Media
• Index: Technology & Engineering

The growth in the use of sensor technology has led to the demand for image fusion: signal processing techniques that can combine information received from different sensors into a single composite image in an efficient and reliable manner. This book brings together classical and modern algorithms and design architectures, demonstrating through applications how these can be implemented. Image Fusion: Algorithms and Applications provides a representative collection of the recent advances in research and development in the field of image fusion, demonstrating both spatial domain and transform domain fusion methods including Bayesian methods, statistical approaches, ICA and wavelet domain techniques. It also includes valuable material on image mosaics, remote sensing applications and performance evaluation. This book will be an invaluable resource to R&D engineers, academic researchers and system developers requiring the most up-to-date and complete information on image fusion algorithms, design architectures and applications. - Combines theory and practice to create a unique point of reference - Contains contributions from leading experts in this rapidly-developing field - Demonstrates potential uses in military, medical and civilian areas

1 Current trends in super-resolution image reconstruction

Antonis Katartzis, Maria Petrou

Communication and Signal Processing Group, Department of Electrical and Electronic Engineering, Imperial College, London, UK

Super-resolution (SR) reconstruction is a branch of image fusion for bandwidth extrapolation beyond the limits of traditional electronic imaging systems. This chapter describes the main principles of SR reconstruction, and provides an overview of the most representative methodologies in the domain. We analyse the advantages and limitations of each set of techniques, present a promising new approach based on Normalised Convolution and robust Bayesian estimation, and perform quantitative and qualitative comparisons using real video sequences.

1.1 Introduction

Super-resolution (SR) is a fusion process for reconstructing a high resolution (HR) image from several low resolution (LR) images covering the same region in the world. It extends classical single frame image reconstruction/restoration methods by simultaneously utilising information from multiple observed images to achieve resolutions higher than that of the original data. These observations can be LR images captured simultaneously or at different times by a single or multiple imaging devices. This methodology, also known as multiframe super-resolution reconstruction, registers the observed images to a common high resolution reference frame in order to formulate the problem of fusion as one of constrained image reconstruction with missing data.

The general strategy that characterises super-resolution comprises three major processing steps [1]:

1. LR image acquisition: Acquisition of a sequence of LR images from the same scene with non-integer (in terms of inter-pixel distances) geometric displacements between any two of the images.

2. Image registration/motion compensation: Estimation of the sub-pixel geometric transformation of each source image with respect to the reference HR desirable grid.

3. HR image reconstruction: Solution of the problem of reconstructing a HR image from the available data supplied by the source images.

The theoretical basis for super-resolution was laid by Papoulis [2], with the Generalised Sampling Theorem. It was shown that a continuous band-limited signal z(x) may be reconstructed from samples of convolutions of z(x) with different filters, assuming these filters satisfy certain conditions. For example, if these filters kill some high frequencies, then there is no unique solution [3]. This is one of the factors that make SR an ill-posed problem. The solution in general does not fulfil Hadamard’s classical requirements of existence, uniqueness and stability: solutions may not exist for all data, they may not be unique (which raises the practically relevant question of identifiability, i.e. the question of whether the data contain enough information to determine the desired quantity), and they may be unstable with respect to data perturbations. The last aspect is very important, since in real-world measurements the presence of noise is inherent. As a consequence, the reconstruction must rely on natural constraints, that is, general a priori assumptions about the physical world, in order to derive an unambiguous output. However, as demonstrated in [4], the quality of reconstruction of the HR image has an upper limit defined by the degree of degradation of the involved LR frames.

This chapter provides a description of the main principles of super-resolution, together with an overview of the most representative methodologies in the domain. We analyse the advantages and limitations of each set of techniques and present a promising new approach based on Normalised Convolution and robust Bayesian...

Table of contents

• Cover image
• Title page
• Table of Contents
• Copyright
• Preface
• List of contributors
• Chapter 1: Current trends in super-resolution image reconstruction
• Chapter 2: Image fusion through multiresolution oversampled decompositions
• Chapter 3: Multisensor and multiresolution image fusion using the linear mixing model
• Chapter 4: Image fusion schemes using ICA bases
• Chapter 5: Statistical modelling for wavelet-domain image fusion
• Chapter 6: Theory and implementation of image fusion methods based on the á trous algorithm
• Chapter 7: Bayesian methods for image fusion
• Chapter 8: Multidimensional fusion by image mosaics
• Chapter 9: Fusion of multispectral and panchromatic images as an optimisation problem
• Chapter 10: Image fusion using optimisation of statistical measurements
• Chapter 11: Fusion of edge maps using statistical approaches
• Chapter 12: Enhancement of multiple sensor images using joint image fusion and blind restoration
• Chapter 13: Empirical mode decomposition for simultaneous image enhancement and fusion
• Chapter 14: Region-based multi-focus image fusion
• Chapter 15: Image fusion techniques for non-destructive testing and remote sensing applications
• Chapter 16: Concepts of image fusion in remote sensing applications
• Chapter 17: Pixel-level image fusion metrics
• Chapter 18: Objectively adaptive image fusion
• Chapter 19: Performance evaluation of image fusion techniques
• Subject index