Vladimir S. Petrovic

Analysis of Features for Rigid Structure Vehicle Type Recognition

We describe an investigation into feature representations for rigid structure recognition framework for recognition of objects with a multitude of classes. The intended application is automatic recognition of vehicle type for secure access and traffic monitoring applications, a problem not hitherto considered at such a level of accuracy. We demonstrate that a relatively simple set of features extracted from sections of car front images can be used to obtain high performance verification and recognition of vehicle type (both car model and class). We describe the approach and resulting system in full, and the results of experiments comparing a wide variety of different features. The final system is capable of recognition rates of over 93% and verification equal error rates of fewer than 5.6% when tested on over 1000 images containing 77 different classes. The system is shown to be robust for a wide range of weather and lighting conditions.

Subjective tests for image fusion evaluation and objective metric validation

This paper focuses on the methodology for perceptual image fusion assessment through comparative tests and validation of objective fusion evaluation metrics. Initially, the theory of subjective fusion evaluation, adopted practice and methods to gauge relevance and significance of individual trials are examined. Further in this context, the methodology, experiences and results of a series of specific, subjective preference tests aimed at relative evaluation of fusion algorithms are presented. Test conditions and experimental procedure are described in detail and a number of explicit fusion metrics derived from the subjective test data are proposed. Relative fusion quality, fusion performance robustness (to content) and personal preference are all assessed by the metrics as different aspects of general image fusion performance. Finally, the methodology for subjective validation of objective fusion metrics using the reported test procedures is presented. In particular, explicit subjective-objective validation algorithms are defined and applied to a range of established objective measures of fusion performance in order to evaluate their subjective relevance.

A unified information-theoretic approach to groupwise non-rigid registration and model building

The non-rigid registration of a group of images shares a common feature with building a model of a group of images: a dense, consistent correspondence across the group. Image registration aims to find the correspondence, while modelling requires it. This paper presents the theoretical framework required to unify these two areas, providing a groupwise registration algorithm, where the inherently groupwise model of the image data becomes an integral part of the registration process. The performance of this algorithm is evaluated by extending the concepts of generalisability and specificity from shape models to image models. This provides an independent metric for comparing registration algorithms of groups of images. Experimental results on MR data of brains for various pairwise and groupwise registration algorithms is presented, and demonstrates the feasibility of the combined registration/modelling framework, as well as providing quantitative evidence for the superiority of groupwise approaches to registration.

Objective pixel-level image fusion performance measure

This paper addresses the issue of objectively measuring the performance of pixel level image fusion systems. The proposed fusion performance metric models the accuracy with which visual information is transferred from the input images to the fused image. Experimental results clearly indicate that the metric is perceptually meaningful.

Objective evaluation of signal-level image fusion performance

An objective measurement framework for signal-level image fusion performance, based on a direct comparison of visual information in the fused and input images, is proposed. The aim is to model and predict subjective fusion performance results otherwise obtained through extremely time- and resource-consuming perceptual evaluation proce- dures. The measure associates visual information with edge, or gradient, information that is initially parametrized at all locations of the inputs and the fused image. A perceptual-information preservation model is then used to quantify the success of information fusion as the accuracy with which local gradient information is transferred from the inputs to the fused image. By considering the perceptual importance of different im- age regions, such local fusion success estimates are integrated into a single, numerical fusion performance score between 0 total information loss and 1 ideal fusion. The proposed metric is optimized and vali- dated using extensive subjective test results and validation procedures. The results clearly indicate that the proposed metric is perceptually meaningful in that it corresponds well with the results of perceptual fu- sion evaluation. Finally, an application of the proposed evaluation ap- proach to fusion algorithm selection and fusion parameter optimization demonstrates its general usefulness.

Groupwise Construction of Appearance Models using Piece-wise Affine Deformations

We describe an algorithm for obtaining correspondences across a group of imagesof deformable objects. The approach is to construct a statistical modelof appearance which can encode the training images as compactly as possible(a Minimum Description Length framework). Correspondences are defined bypiece-wise linear interpolation between a set of control points defined oneach image. Given such points a model can be constructed, which can approximateevery image in the set. The description length encodes the cost of the model,the parameters and most importantly, the residuals not explained by the model.By modifying the positions of the control points we can optimise the descriptionlength, leading to good correspondence. We describe the algorithm in detailand give examples of its application to MR brain images and to faces. We alsodescribe experiments which use a recently-introduced specificity measureto evaluate the performance of different components of the algorithm.

Computing Accurate Correspondences across Groups of Images

Groupwise image registration algorithms seek to establish dense correspondences between sets of images. Typically, they involve iteratively improving the registration between each image and an evolving mean. A variety of methods have been proposed, which differ in their choice of objective function, representation of deformation field, and optimization methods. Given the complexity of the task, the final accuracy is significantly affected by the choices made for each component. Here, we present a groupwise registration algorithm which can take advantage of the statistics of both the image intensities and the range of shapes across the group to achieve accurate matching. By testing on large sets of images (in both 2D and 3D), we explore the effects of using different image representations and different statistical shape constraints. We demonstrate that careful choice of such representations can lead to significant improvements in overall performance.

Objective image fusion performance characterisation

Image fusion as a way of combining multiple image signals into a single fused image has in recent years been extensively researched for a variety of multisensor applications. Choosing an optimal fusion approach for each application from the plethora of algorithms available however, remains a largely open issue. A small number of metrics proposed so far provide only a rough, numerical estimate of fusion performance with limited understanding of the relative merits of different fusion schemes. This paper proposes a method for comprehensive, objective, image fusion performance characterisation using a fusion evaluation framework based on gradient information representation. The method provides an in-depth analysis of fusion performance by quantifying: information contributions by each sensor, fusion gain, fusion information loss and fusion artifacts (artificial information created). It is demonstrated on the evaluation of an extensive dataset of multisensor images fused with a wide range of established image fusion algorithms. The results demonstrate and quantify a number of well known issues concerning the performance of these schemes and provide a useful insight into a number of more subtle yet important fusion performance effects not immediately accessible to an observerImage fusion is finding increasing application in areas such as medical imaging, remote sensing or military surveillance using sensor networks. Many of these applications demand highly compressed data combined with error resilient coding due to the characteristics of the communication channel. In this respect, JPEG2000 has many advantages over previous image coding standards. This paper evaluates and compares quality metrics for lossy compression using JPEG2000. Three representative image fusion algorithms: simple averaging, contrast pyramid and dual-tree complex wavelet transform based fusion have been considered. Numerous infrared and visible test images have been used. We compare these results with a psychophysical study where participants were asked to perform specific tasks and assess image fusion quality. The results show that there is a correlation between most of the metrics and the psychophysical evaluation. They also indicate that selection of the correct fusion method has more impact on performance than the presence of compression.

Assessing the accuracy of non-rigid registration with and without ground truth

We compare two methods for assessing the performance of groupwise non-rigid registration algorithms. The first approach, which has been described previously, utilizes a measure of overlap between ground-truth anatomical labels. The second, which is new, exploits the fact that, given a set of nonrigidly registered images, a generative statistical model of appearance can be constructed. We observe that the quality of this model depends on the quality of the registration, and define measures of model specificity and generalisation - based on comparing synthetic images sampled from the model, with those in the original image set - that can be used to assess model/registration quality. We show that both approaches detect the loss of registration accuracy as the alignment of a set of correctly registered MR images of the brain is progressively perturbed. We compare the sensitivities of the two approaches and show that, as well as requiring no ground truth, specificity provides the most sensitive measure of misregistration

Objectively adaptive image fusion

Signal-level image fusion has been the focus of considerable research attention in recent years with a plethora of algorithms proposed, using a host of image processing and information fusion techniques. Yet what is an optimal information fusion strategy or spectral decomposition that should precede it for any multi-sensor data cannot be defined a priori. This could be learned by either evaluating fusion algorithms subjectively or indeed through a small number of available objective metrics on a large set of relevant sample data. This is not practical however and is limited in that it provides no guarantee of optimal performance should realistic input conditions be different from the sample data. This paper proposes and examines the viability of a powerful framework for objectively adaptive image fusion that explicitly optimises fusion performance for a broad range of input conditions. The idea is to employ the concepts used in objective image fusion evaluation to optimally adapt the fusion process to the input conditions. Specific focus is on fusion for display, which has broad appeal in a wide range of fusion applications such as night vision, avionics and medical imaging. By integrating objective fusion metrics shown to be subjectively relevant into conventional fusion algorithms the framework is used to adapt fusion parameters to achieve optimal fusion display. The results show that the proposed framework achieves a considerable improvement in both level and robustness of fusion performance on a wide array of multi-sensor images and image sequences.