Fusion performance using a validation approach

Abstract

The evaluation ofimagefusion performance is an active area of research with a variety ofdiferent approaches under investigation.Examples of techniques include those which aim to evaluate the quality offused images for human display and are based on perception metrics, and others which utilise standard image metrics to measure edge densities, noise and other such characteristics. This latter approach can produce a good performance estimate under ideal conditions but starts to break-down in high noise environments, for example. A novel solution is to use image validation metricsfor the evaluation. Image validation metrics are concerned more with the differences between images than the absolute pixel values and as such should exhibit greater resilience to changing environmental conditions. This paper furthers existing work in this area to include a family of validation metrics for image fusion and draws upon existing image validation tools to expedite the process. Assessment of fusion performance is carried out between diferent fusion architectures over a range of trials and different cameras. Comparisons offused images to the source imagery are made and consideration is also given to the temporalperformance ofthe methodpresented.

Towards cognitive image fusion

“…However, this usually means time consuming and often expensive experiments involving a large number of human subjects. In recent years, a number of computational image fusion quality assessment metrics have therefore been proposed [2,3,[5][6][7][12][13][14]36,42,44,46,49,[52][53][54][55]. Although some of these metrics agree with human visual perception to some extent, most of them cannot predict observer performance for different input imagery and scenarios.…”

Applied Multi-Dimensional Fusion

“…An alternative family of measures is the set of image validation measures, which calculate the difference (or similarity) between two images for a given characteristic [14]. One or more of the original input modalities is chosen as a reference when assessing output from an image fusion technique using image validation measures.…”

Mutual Information Metric Evaluation for PET/MRI Image Fusion

“…As part of the general theme of fusion evaluation for such things as user acceptance [18], target recognition [19], and mission effectiveness, there is a growing interest to develop methods that address the scored performance [20] of image fusion algorithms. Angell, [21] presented some image validation metrics of signal-to-noise (SNR) ratio, variance, and edge density; that were combined into cross-correlation, image quality, structural similarity, and peak SNR metrics. Angell applied these metrics over a host of standard image processing routines such as the Laplacian and wavelets.…”