EVALUATING THE SENSITIVITY OF IMAGE FUSION QUALITY METRICS TO IMAGE DEGRADATION IN SATELLITE IMAGERY F. Samadzadegan, F. DadrasJavan* Dept. of Surveying and Geomatics, University College of Engineering, University of Tehran, Tehran, Iran, North Amir abaad, University College of Engineering, University of Tehran, Tehran,Iran Tel: +98 21 88008841 Fax: +98 21 88008837 E-mail:{samadz,fdadrasjavan}@ut.ac.ir Key Words: Image Fusion, Quality Assessment, Quality Metrics, High Resolution Satellite Imagery Abstract -Referring to the high potential of topographic satellites in collecting high resolution panchromatic imagery and high spectral, multi spectral imagery, the purpose of image fusion is to produce a new image data with high spatial and spectral characteristics. It is necessary to evaluate the quality of fused image by some quality metrics before using this product in later applications. Up to now, several metrics have been proposed for assessment of image quality; which can also be applied for quality evaluation of fused images. However, it seems more investigations are needed to inspect potential evaluation of image fusion quality metrics to registration accuracy, especially in high resolution satellite imagery. This paper focuses on such studies and, using different image fusion quality metrics, experiments conducted to evaluate the sensitivity of them on high resolution satellite imagery in an urban area. The obtained results clearly reveal that these metrics sometimes do not behave robust in the whole area and also their achieved results are inconsistence in different patch areas in comparison with the whole image. 1. INTRODUCTION Nowadays, as many earth observation satellites provide both high resolution panchromatic and low resolution multispectral images; image fusion plays an important role in the field of remote sensing to produce a high resolution multispectral image (Ranchin and Wald, 2000). Image fusion has become very important in many applications of remote sensing like land use classification, detecting changes, updating maps, monitoring hazards and many other Geo-information applications (Reyes et al., 2004; Ehlers et al., 2008). Nevertheless, there are significant color distortions in processed image achieved through fusion due to the registration accuracy. So, quality assessment of these data is crucial before using them in other next process of object extraction or recognition (Ehlers et al., 2008). As no reference multispectral images are available in real applications and visual evaluation of processed images is time consuming task and dependent on expert knowledge, lots of efforts in the field are directed to objective quality assessment of fused images based on the initial concepts of image quality metrics (Thomas and Wald, 2006a). Many image quality assessment algorithms have been shown to behave consistently when applied to distorted images created from the same original image, using the same type of radiometric and spectral characteristics. However, the effectiveness of these models degrades significantly when applied to a set of images originating from different reference images, and/or including a variety of different types of distortion. This study focuses on capability evaluation of different image fusion quality metrics which could be applied on assessment of high resolution satellite imagery. The mentioned strategies are developed to inspect the quality of pan sharpening QuickBird panchromatic and multi spectral images. 2. IMAGE FUSION QUALITY ASSESSMENT METRICS The goal of image fusion techniques is to combine, preserve and present all of the important input spectral and spatial information in a single output image. In most applications, quality of the fused images has primary importance. Available Image Fusion Quality Metrics (IFQMs) fall into two categories: Subjective assessments by humans and Objective assessment by algorithms designed to mimic human subjectivity. While subjective assessment is the ultimate gauge of image quality assessment, it is time-consuming and cumbersome. Thus objective approaches which predict subjective image quality accurately and robustly are of considerable value in remote sensing community (Thomas and Wald, 2006a).