Fusion of Insar High Resolution Imagery and Low Resolution Optical Imagery

High resolution (1.25 meter) ortho-rectified radar imagery (ORI) from an airborne InSAR platform was recently created for the whole of Great Britain as part of the NEXTMapUK program which simultaneuosly generated a DSM (Digital Surface Model) and DTM (Digital Terrain Model) with 5 meter samples, for the same national area. A similar initiative is underway in the USA. In order to provide a broader range of useful products for the user community, the NEXTMap data-base, which normally includes a single X-Band channel ORI (OrthoRectified Image) along with associated DSM and DTM, will be supplemented by an optical layer. The optical layer may be able to take advantage of relatively recent air-photo archival data in some urban areas, but in order to economically capture large rural areas it is has been proposed that widely available, lower-cost, coarse-resolution multi-spectral (MS) satellite imagery from candidate sources such as SPOT (10 m), ASTER (15 m) or Landsat(30 m) be merged with the 1.25 meter STAR-3i ORIs being created currently. Thus a new high-resolution hybrid color layer would be generated. This paper addresses some of the challenges associated with this task. There are basically three technical issues that need to be resolved in this process: (1) the candidate satellite imagery must be orthorectified, (2) the MS optical and radar imagery must be merged despite pixel size differences of a factor of 8, 12 or 24, depending on the MS source, and (3) color transformations need to be considered either preor post-merge. The least onerous problem is the ortho-rectification of the optical imagery, since the radar ORI is a source of ground control points and there is a simultaneous DSM available. The merging activity is more challenging since it is desired to create a stable and predictable process in which the spectral content of the MS is largely preserved in the merged product. A modified version of a wavelet transformation process has been developed at UNB and applied to this problem. Lastly, there are color issues because while it is desireable to create a facsimile of the standard RGB color imagery associated with aerial photography, at least two of the MS source candidates (ASTER and SPOT) have spectral bands that are essentially Green, Red and Near-IR. Several approaches have been examined. In this paper we will describe these issues, the processes evolved and the results of tests on the candidate MS images.