ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19 – 30 www.elsevier.com/locate/isprsjprs

Fusion of hyperspectral and radar data using the IHS transformation to enhance urban surface features C.-M. Chen a,*, G.F. Hepner b, R.R. Forster b a

Department of Geography, National Taiwan Normal University, Taipei, Taiwan, ROC b Department of Geography, University of Utah, Salt Lake City, USA Received 28 June 2002; accepted 31 January 2003

Abstract The Intensity – Hue – Saturation (IHS) transformation is used to integrate the high spectral resolution, provided by hyperspectral data (Airborne Visible Infrared Imaging Spectrometer, AVIRIS), and the surface texture information, derived from radar data (Topographic Synthetic Aperture Radar, TOPSAR), into a single image of an urban area. This transformed image is superimposed on the Digital Elevation Model (DEM) data derived from TOPSAR data to create a 3D perspective view. The ambiguity of several urban land cover types is resolved to a larger degree using the higher spectral and spatial resolutions and the synergistic visual content provided by the fused image in a 3D perspective. For urban areas that are at risk from geological hazards (e.g., avalanches, mudflows, and debris flows), the fused image provides a cost-effective product, rich in the information necessary for assessment and mitigation of these hazards. D 2003 Elsevier Science B.V. All rights reserved. Keywords: hyperspectral; SAR; AVIRIS; TOPSAR; IHS transformation; image fusion; urban hazards

1. Introduction The data fusion of multisensor data has received tremendous attention in the remote sensing literature (Yao and Gilbert, 1984; Welch and Ehlers, 1988; Chavez et al., 1991; Weydahl et al., 1995; Niemann et al., 1998; Saraf, 1999; Zhang, 1999; Gamba and Houshmand, 1999). For example, NASA’s Jet Propulsion Laboratory merged three datasets to create a 3D perspective view of Pasadena area, California in which the Shuttle Radar Topography Mission (SRTM) supplied the elevation data, Landsat TM provided the * Corresponding author. Tel.: +886-2-23637874x119; fax: +886-2-23691770. E-mail address: [email protected] (C.-M. Chen).

land surface color, and USGS digital aerial photography supplied the image details. The resultant image demonstrated that the advantages of the spatial detail, the spectral detail, and the topographical detail derived from different sources could be maximized by a data fusion technique (NASA, 2003). In an urban area, many land cover types/surface materials are spectrally similar. This makes it extremely difficult to analyze an urban scene using a single sensor with a limited spectral range (Forster, 1985; Hepner et al., 1998). For example, some land cover types are spectrally indistinguishable from each other within a Landsat TM scene, such as asphalt vs. water or shadow, trees and shrubs vs. lawns, and bare soil vs. newly completed concrete sections (Wheeler, 1985). Some of these features can be discriminated in a radar image

0924-2716/03/$ - see front matter D 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0924-2716(03)00014-5

20

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

based on their dielectric properties and surface roughness. For instance, building walls oriented orthogonal to the radar look direction form corner reflectors and have relatively strong signal returns. A smooth surface of bare soil, which acts as specular reflector, will result in relatively low signal returns. However, trees can introduce interpretation uncertainty by producing bright returns similar to buildings. Wet soil and other urban features with high dielectric constants (e.g., vegetation, metal roofs) are confused in a radar image (Henderson and Xia, 1999). Data fusion is capable of integrating different imagery data creating more information than can be derived from a single sensor. Geophysical information extracted from the optical and microwave wavelengths has proven to be useful for characterization of the chemical composition and morphology of surfaces (Kierein-Young, 1997). This research measured the surface– radiation interactions at the molecular scale giving information about the mineralogical makeup of the surface. Analysis of the transmitted radiation scattered from the surface textural and dielectric variations provided information about the morphology and moisture content of the surface. The integrated image, therefore, provides complementary information and improves the understanding of the compositional variability in the surfaces. If one adds geometrical information derived from the imagery data and all are portrayed using modern visualization techniques, a very powerful characterization of the urban area can be made solely from imagery sources. The objective of our study is to demonstrate a rapid and effective approach for merging Airborne Visible Infrared Imaging Spectrometer (AVIRIS) and Topographic Synthetic Aperture Radar (TOPSAR) data. An image-enhancement technique, Intensity– Hue – Saturation (IHS) transformation, is adopted to conduct the data fusion. The fusion provides a transformed image with significant improvement in the visual discrimination of features.

2. Study area and data acquisition 2.1. Study area The urban area of Park City, Utah (Fig. 1) is the study area selected for this study. Park City contains

the diversity of land cover types and surface materials that can be found in many western US cities. Many common vegetation communities such as Douglas fir, lodgepole pine, limber pine, ponderosa pine, white fir, blue spruce, and aspen are also found within the study area as introduced species, or naturally at higher elevations (Murphy, 1981). The elevation of this highland area ranges from 2076 (residential area) to 2657 m (mountain area). It provides a typical example of where rapid urbanization is spreading from lowlands to adjacent steep uplands creating hazardous situations such as avalanches, mudflows, and debris flows. Results from this analysis should have application to other rapidly growing cities in mountainous, semi-arid regions of the world. 2.2. TOPSAR data product The airborne synthetic aperture radar (SAR) data product used in this study was acquired by the NASA/ JPL AIRSAR in TOPSAR mode (AIRSAR, 2003). The TOPSAR mode selected (XTI1) generated a C-band interferometric Digital Elevation Model (DEM), a Cband VV image (transmit and receive vertical polarization) along with L- and P-band fully polarimetric images. The integrated TOPSAR data product also consists of derived data types such as an incidence angle map and a C-band interferometric correlation map. All data files are in ground range projection and in range line format in which each record in the data file corresponds to constant along-track position (azimuth) and varying across-track position (range). The ground data projection is conducted using the DEM data derived from the C-band interferometry. The DEM data represent the elevation of the terrain above a spherical approximation to the WGS-84 ellipsoid with a 10-m pixel size. The data elevations agree to the map elevations (Draper and Dromedary Peak, 1:24,000 scale maps) within 3 –8 m. The TOPSAR data over Park City area were collected on 28 October 1998. The P-band data were collected at 20-MHz bandwidth corresponding to 6.7m resolution in the range direction. The C- and Lband data were collected at 40 MHz yielding a 3.3-m resolution. The resultant P-band data are in the slant range and are not co-registered with the other two frequencies due to difference in spatial resolution resulting from the wider bandwidth.

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

21

Fig. 1. The AVIRIS scene of Park City, Utah, US. The white box indicates the area to be merged with TOPSAR data and the white marks indicate the ground control points for georectification.

2.3. AVIRIS data AVIRIS is an optical sensor that records the upwelling spectral radiance in 224 contiguous spectral bands with wavelengths from 400 to 2500 nm. The high-altitude AVIRIS data of the Park City area have a pixel size of 20 m and were obtained on 5 August 1998. The data range for this imagery is from Band 1, with a bandcenter of 369.07 nm, to Band 224, with a bandcenter of 2507.50 nm. The bandwidth is approximately 10 nm. The AVIRIS radiance data of the study site are calibrated to reflectance data using a radiative transfer model (ATREM) and in situ spectra from several field calibration sites. 2.4. Ground truth data The US Geological Survey 1:24,000 scale Digital Orthophotoquad (DOQ) maps of the Park City area

are used to provide ground control points and to geometrically rectify the TOPSAR and AVIRIS images. Since the DOQ maps were obtained in 1993, whereas the AVIRIS and the TOPSAR data were acquired in 1998, many land cover types have changed in the study area. Field checks were conducted to provide additional ground truth data.

3. Techniques for data fusion 3.1. Minimum noise fraction (MNF) transformation There are 224 bands in the AVIRIS data. However, only three bands can be used to perform the IHS transformation. Neighbouring spectral bands exhibit high interband correlations resulting in a large amount of redundancy. Therefore, the MNF transformation is used to reduce the data dimensionality of AVIRIS data before performing the data fusion.

22

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

The MNF transformation was first developed as an alternative to the principal components analysis (PCA) for the Airborne Thematic Mapper (ATM) 10-band sensor (Green et al., 1988). The MNF transformation is defined as a two-step cascaded PCA (Research Systems Inc., 1998). The first step, based on an estimated noise covariance matrix, is to decorrelate and rescale the data noise, where the noise has unit variance and no band-to-band correlations. The next step is a standard PCA of the noise-whitened data. The MNF transformation was derived as an analogue of the PCA and has all the properties of the PCA, including the primary characteristic of optimally concentrating the information content of the data in as small a number of components as possible (Lee et al., 1990). This transformation is equivalent to principal components when the noise variance is the same in all bands (Green et al., 1988). By applying the MNF to the Geophysical and Environment Research (GER) 64-band data, the results demonstrated the effectiveness of this transformation for noise adjustment in both the spatial and spectral domains (Lee et al., 1990). The MNF transformation is chosen for this study because it not only can reduce the data dimensionality of the AVIRIS data, but also order components in terms of image quality, the signal–noise ratio (SNR). Although PCA can efficiently compress hyperspectral data into a few components, this high concentration of total covariance only in the few first components inevitably results from part of the noise variance (Green et al., 1988; Lee et al., 1990; Chen, 2000). The special capability of the MNF, which shows gradually decreasing image quality with increasing component number, cannot be accomplished easily using other techniques, such as PCA and factor analysis.

extracted from the composite SIR-B/TM image exceeds those obtained from individual TM and SIR-B images by approximately 10% and 25%, respectively. The general IHS procedure uses three bands of a lower spatial resolution dataset and transforms these data to IHS space. A contrast stretch is then applied to the higher spatial resolution image so that the stretched image has approximately the same variance and average as the intensity component image. Then, the stretched, higher resolution image replaces the intensity component before the image is transformed back into the original color space (Chavez et al., 1991). The IHS color coordinate system is based on a hypothetical color sphere. The vertical axis represents intensity, which ranges from 0 (black) to 255 (white). The circumference of the sphere represents hue, which is the dominant wavelength of color. Hue ranges from 0 at the midpoint of red tones through green, blue and back to 255, adjacent to 0. Saturation represents the purity of the color and ranges from 0 at the center of the color sphere to 255 at the circumference (Jensen, 1996). The IHS values can be derived from the RGB values through transformation equations. The following equations are used to compute IHS values for a RGB image (BV1, BV2, and BV3). The IHS values can also be converted back into RGB values using the inverse of these equations (Pellemans et al., 1993). I ¼ ðBV1 þ BV2 þ BV3 Þ=3 H ¼ ½arctanð2BV1  BV2  BV3 Þ=M3ðBV2  BV3 Þ þC ð1Þ S ¼ M6ðBV21 þ BV22 þ BV23  BV1 BV2  BV1 BV3

3.2. IHS color transformation IHS is one of the most often used methods to merge multisensor image data. It was first used to merge Landsat MSS with Return Beam Vidicon (RBV) data and Landsat MSS with Heat Capacity Mapping Mission data (Haydn et al., 1982). The IHS transformation has also been used to merge SIR-B and Landsat TM images (Welch and Ehlers, 1988). It was found that the completeness of cartographic features

 BV2 BV3 Þ0:5 =3 where, C = 0, if BV2 z BV3; C = k, if BV2 < BV3. Since the objective of this study is to demonstrate a rapid and effective approach for merging AVIRIS data and TOPSAR data, a color transformation technique such as IHS meets the criteria. It can enhance image features, improve spatial resolution, and integrate disparate data at low processing cost.

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

Other color-related techniques such as color composites (RGB) and luminance – chrominance (YIQ) do not have all these advantages (Pohl and van Genderen, 1998). For multidimensional data with more than three spectral bands, the selection of an adequate threeband composite in RGB color system is crucial for maximizing the spectral discriminability in the IHS transformation. In a study using IHS transformation of Landsat-5 TM data for burned land mapping, the burned area was best discriminated by transforming the three bands TM7– TM4 –TM1 corresponding to red, green, blue color planes to the IHS color system (Koutsias et al., 2000). In our study, the reduction of the dimensionality and the compression of the spectral information in the AVIRIS data are accomplished by the MNF transformation. The MNF components with most spectral information are used to determine the three-band composite in RGB color system.

23

4. Image processing 4.1. Image georectification The USGS 1:24000 scale DOQ maps are used as the reference images. The study area is covered by two map sheets: West Park City, Utah and East Park City, Utah. These images are mosaicked to create a single map. Since the 1998 AVIRIS flight line over the study area was from south to north, the AVIRIS image was first rotated by 180j to match the orientation of the DOQ maps and to facilitate the selection of ground control points (GCPs). The image-to-map rectification process is used to register the AVIRIS image to the reference map. The AVIRIS image of the study area is relatively small (8  8 km) and some GCPs can be easily found at road intersections in this urban scene. Fifteen GCPs are selected in the image (Fig. 1) along with their corresponding map coordinates. A 1st-order polynomial and the nearest neigh-

Fig. 2. The rotated and georectified TOPSAR L-band HV image of the study area.

24

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

Fig. 3. The first nine components of the MNF-transformed AVIRIS image.

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30 Table 1 Eigenvalues of the first nine MNF components Component 1

2

3

4

5

6

7

8

9

Eigenvalue 4100 3166 2318 1438 758 545 371 316 198

bor interpolation are chosen as the warping and gray level resampling methods, respectively. The output projection for the registered image is UTM zone 12 with NAD83 datum and the AVIRIS image is resampled to 10-m pixel size from the original 20-m pixels. The total RMS error is 0.76 pixel. Similar geometric corrections are applied to the TOPSAR data (Fig. 2). The four linearly polarized Lband images (HH, VV, HV, and TP (Total Power)) and the DEM (derived from C-band interferometry) are rotated 270j to the same orientation as the reference map. Fifteen GCPs were measured. The pixel size of TOPSAR images is resampled to 10 m when they are delivered from JPL. The same pixel size was used during the georectification. It is more difficult to

25

accurately identify GCPs in the TOPSAR image than in the AVIRIS image because of radar speckle and imaging geometry. Thus, the RMS error of the TOPSAR rectification is 1.20 pixels, which is higher than that of the AVIRIS data. The fit between the georeferenced AVIRIS and TOPSAR images is validated by measuring well distributed common points. The AVIRIS image agrees to the TOPSAR image within two pixels. 4.2. Imaging processing for the data fusion As the goal of this study is to merge the AVIRIS and TOPSAR image data with the TOPSAR DEM, the following procedures are applied. The MNF transformation was performed to reduce the dimensionality of the AVIRIS data. Among the first 9 components, components 1, 2, and 3 revealed most of the surface features, so they were chosen to be converted to the IHS space. The intensity range of the AVIRIS data was converted from 16 to 8 bits to match the range of

Fig. 4. The color composite image of the MNF components 2 (R), 1 (G), and 3(B). The red box indicates the area to be merged with TOPSAR data.

26

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

the hue in the IHS color system. The first three MNF components of the AVIRIS data were assigned to red (band 2), green (band 1), and blue (band 3), where urban surface features have strong color contrast (e.g., built-up areas in magenta color; woodlands in cyan color; and grass in yellow color), and they were then converted into the IHS color system (Fig. 4). By visually comparing the L-band HH, VV, HV-polarized, and C-band VV-polarized images, the HV image shows more contrast than the other ones for detecting urban surface features. Thus, the L-band HV TOPSAR image was chosen as the high spatial resolution image to replace the intensity of the IHS-transformed AVIRIS image. The number of looks processed in azimuth was limited at 36 to prevent the further degradation of the spatial detail in the TOPSAR Lband HV image. A contrast stretch algorithm, the square root stretch, is applied to the L-band HV image so that it has approximately the same variance as the intensity component image. The intensity component of the IHS AVIRIS image is replaced by the contrast-

stretched TOPSAR L-band HV image. The IHS-transformed image is converted back to RGB color space. The merged image is superimposed on the DEM data to create a 3D perspective view of the study area.

5. Results As expected, the MNF successfully ordered components in terms of image quality (Fig. 3). There is a definite trend of increasing noise (which is correlated to the eigenvalues) with increasing component number (Table 1). The grass/crop vegetation has higher brightness values than those of built-up areas in the first component. The grassland (golf course) in the upper portion of the second component is readily identifiable due to very high brightness, and the ponds in the golf courses are relatively dark. The bare soil of ski trails and mine tailings is revealed in the third component with very high brightness. Through comparison with the DOQ map, it was found that the

Fig. 5. The contrast-stretched, L-band HV TOPSAR image.

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

magenta color indicated buildings, roads, and mining tailings. The cyan color revealed mature woodlands. The yellow color depicted grass and crops. The blue color indicates bare soil. The green spots surrounded by yellow color are water bodies (ponds). The north sides of mountains are in shadow and depicted as green in the lower portion of the image. In the contrast-stretched L-band HV TOPSAR image (Fig. 5), ‘‘very bright’’ areas (high radar return) included buildings and mature woodlands; ‘‘very dark’’ areas (low radar return) included ski trails, roads, and grasslands (ranches and golf courses). It is difficult to separate buildings from mature woodlands or ski trails from roads based on the brightness as they have similar radar return intensities. However, several of these types could be separated by their different surface roughness and texture. For example,

27

the central city areas have a high degree of speckle due to surface roughness variations and multiple path reflections, but the woodlands appear to have significantly less texture. If ground truth information or previous knowledge about the study area was not available, the highways would not be separable from the ski trails because both have similar brightness and texture in the L – HV image. Compared to the original MNF-transformed AVIRIS image, the spatial resolution of the AVIRIS/ TOPSAR merged image is significantly enhanced by the L-band HV TOPSAR image (Fig. 6). For example, the linear features such as roads and ski trails are much more identifiable than those in the original AVIRIS image. The merged image also revealed differences in surface roughness, which are not visible in the AVIRIS image. Originally, the buildings and

Fig. 6. The merged image of the MNF-transformed AVIRIS scene and the TOPSAR.

28

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

mining tailings were spectrally indistinguishable in the MNF-transformed AVIRIS image because both of them were indicated by the same color, blue (Fig. 4). In the merged image (Fig. 6), the buildings have a very high reflectance (brighter) and the mine tailings a very low one (darker). On the other hand, the original L-band HV TOPSAR image is spectrally enhanced by the AVIRIS image. For example, both pine trees and buildings have very high amplitude in the L-band HV TOPSAR image (Fig. 5). However, they are separated by the different colors in the merged image (Fig. 6). The pine trees are indicated by cyan and green (in shadow), and the buildings are represented by magenta. It should be noted that the spectral resolution of the AVIRIS MNF image was partially degraded by the Lband HV TOPSAR image. Several surface features (e.g., ski trails, fairways of the golf courses, and lawns of the softball field and the city cemetery) have very low intensities due to relatively smooth surfaces.

When the intensity of the AVIRIS MNF image is replaced by the L-band HV TOPSAR image and converted back to the RGB space, these surface features are too dark to differentiate from each other and loose their spectral information. A 3D perspective view generated by superimposing the AVIRIS/TOPSAR merged image on the DEM is shown in Fig. 7. No vertical exaggeration is applied to this image. The Park City ski area is shown on the right portion of the image. The downtown area of Park City is shown in the middle as a magenta color belt. The dominant vegetation community, Gambel oak (Quercus gambelii), is found at lower elevations around the left side of the Park City downtown area and indicated by a gray color. Pine trees are also found within the study area at higher elevations and represented by cyan and green. The confusion between roads and ski trails in the L-band HV TOPSAR image is resolved by the topography and the spatial pattern in the topographical perspective view.

Fig. 7. The 3D perspective view of the study area (180j rotated).

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

6. Conclusions In this study, the merged image and its 3D perspective view demonstrate that the IHS transformation is capable of integrating AVIRIS and TOPSAR images. Urban surface features are more distinguishable from each other using the AVIRIS/TOPSAR merged image than by using a single sensor data. The spatial resolution derived from the L-band HV TOPSAR image successfully enhanced the AVIRIS image. On the other hand, the spectral resolution derived from the MNF-transformed AVIRIS image also improved the discriminability in the L-band HV TOPSAR image. As expected, the spectral confusion between buildings and mining tailings in the AVIRIS image is resolvable by the different intensities in the TOPSAR L-band HV image due to differences in surface roughness and multipath reflections. The confusion between trees and buildings, due to the high intensities in the radar image, was resolved by the different colors generated by the MNF-transformed AVIRIS image. Moreover, the 3D perspective revealed topographic and textural information, which resolved the confusion between the ski trails and highways in the TOPSAR image. The major source of color distortion in the merged image is the result of the limited contrast of the TOPSAR L-band HV image. Several surface features such as fairways of golf courses and ski trails were not differentiable in the TOPSAR image due to the low reflectance close to 0. The low intensity resulted in low RGB values for these features in the merged image. Consequently, the color (i.e. spectral resolution) of these features was lost. The integrated use of the MNF transformation, IHS transformation, and DEM visualization overlay provides rapid, easy, and effective image fusion of AVIRIS and TOPSAR data for urban areas. The various output products can be used in a variety of analytical procedures to extract detailed information on urban surface materials. The same procedure is also applicable to spaceborne multispectral sensors such as Landsat TM and SPOT, which are more readily available than airborne AVIRIS and TOPSAR. However, the information content in the hyperspectral data is much richer than that from multispectral data. The advantage of hyperspectral data over multispectral data for the use of visual enhancement should be

29

further investigated. In this study, hyperspectral data are simply used to provide three components for visual enhancement. For future studies, an imaging spectroscopy approach, which uses contiguous bands for the direct identification of materials with diagnostic absorption features, can be used to extract more urban surface features than the MNF transformation. The individual surface features derived from hyperspectral data can be combined into specific land cover classes to meet the data requirements of various urban applications.

Acknowledgements This study is supported by the Taiwan National Science Council (NSC 91-2415-H-003-005).

References AIRSAR (JPL/NASA), 2003. TOPSAR data operation mode. http:// airsar.jpl.nasa.gov/documents/dataformat.htm (accessed 25 January 2003). Chavez, P.S., Sides, S.C., Anderson, J.A., 1991. Comparison of three different methods to merge multiresolution and multispectral data: Landsat TM and SPOT Panchromatic. Photogrammetric Engineering and Remote Sensing 57 (3), 295 – 303. Chen, C.-M., 2000. Comparison of principal components analysis and minimum noise fraction transformation for reducing the dimensionality of hyperspectral imagery. Geographical Research 33 (1), 163 – 178. Forster, B.C., 1985. An examination of some problems and solutions in monitoring urban areas from satellite platforms. International Journal of Remote Sensing 6 (1), 139 – 151. Gamba, P., Houshmand, B., 1999. Three-dimensional road network by fusion of polarimetric and interferometric SAR data. Proceedings of IGARSS’99 I, 302 – 304. Green, A.A., Berman, M., Switzer, P., Craig, M.D., 1988. A transformation for ordering multispectral data in terms of image quality with implications for noise removal. IEEE Transactions on Geoscience and Remote Sensing 26 (1), 65 – 74. Haydn, R., Dalke, G.W., Henkel, J., Bare, J.E., 1982. Application of the IHS color transform to the processing of multisensor data and image enhancement. Proceedings of the International Symposium on Remote Sensing of Arid and Semi-Arid Lands, Cairo, Egypt, pp. 599 – 616. Henderson, F.M., Xia, Z.G., 1999. Chapter 15: radar applications in urban analysis, settlement detection and population estimation, 3rd ed. Manual of Remote Sensing, vol. 2. Wiley, New York, pp. 759 – 761. Hepner, G.F., Houshmand, B., Kulikov, I., Bryant, N., 1998. Investigation of the Integration of AVIRIS and IFSAR for urban

30

C.-M. Chen et al. / ISPRS Journal of Photogrammetry & Remote Sensing 58 (2003) 19–30

analysis. Photogrammetric Engineering and Remote Sensing 64 (8), 813 – 820. Jensen, J.R., 1996. Merging different types of remotely sensed data for effective visual display. Introductory Digital Image Processing: A Remote Sensing Perspective, 2nd ed. Prentice-Hall, Upper Saddle River, NJ, pp. 100 – 103. Kierein-Young, K.S., 1997. The integration of optical and radar data to characterize mineralogy and morphology of surfaces in Death Valley, California, U.S.A. International Journal of Remote Sensing 18 (7), 1517 – 1541. Koutsias, N., Karteris, M., Chuvieco, E., 2000. The use of intensity – hue – saturation transformation of Landsat-5 Thematic Mapper data for burned land mapping. Photogrammetric Engineering and Remote Sensing 66 (7), 829 – 839. Lee, J.B., Woodyatt, A.S., Berman, M., 1990. Enhancement of high spectral resolution remote sensing data by a noise-adjusted principal components transform. IEEE Transactions on Geoscience and Remote Sensing 28 (3), 295 – 304. Murphy, D.R., 1981. Vegetation zones. Atlas of Utah. Brigham Young University, Provo, UT, pp. 31 – 32. NASA, 2003. SRTM Web Page. http://www.jpl.nasa.gov/srtm (accessed 25 January 2003). Niemann, K.O., Goodenough, D.G., Marceau, D., Hay, G., 1998. A practical alternative for fusion of hyperspectral data with high resolution imagery. Proceedings of IEEE International Geoscience and Remote Sensing Symposium 1, 174 – 176. Pellemans, A.H., Jordans, R.W., Allewijn, R., 1993. Merging multispectral and panchromatic SPOT images with respect to the

radiometric properties of the sensor. Photogrammetric Engineering and Remote Sensing 59 (1), 81 – 87. Pohl, C., van Genderen, J.L., 1998. Multisensor image fusion in remote sensing: concepts, methods and applications. International Journal of Remote Sensing 19 (5), 823 – 854. Research Systems Inc., 1998. Minimum Noise Fraction (MNF). Exploring ENVI—an RSI Training Course for ENVI 3.0, pp. 111 – 114. Saraf, A.K., 1999. IRS – 1C – LISS – III and PAN data fusion: an approach to improve remote sensing based mapping techniques. International Journal of Remote Sensing 20 (10), 1929 – 1934. Welch, R., Ehlers, M., 1988. Cartographic feature extraction with integrated SIR-B and Landsat TM images. International Journal of Remote Sensing 9 (5), 873 – 889. Weydahl, D.J., Becquey, X., Tollefsen, T., 1995. Combining ERS-1 SAR with optical satellite data over urban areas. Proceedings of the IEEE International Geoscience and Remote Sensing Symposium 3, 2161 – 2163. Wheeler, D.J., 1985. Evaluation of thematic mapper data for determining urban land cover. PhD dissertation, University of Utah. Department of Geography, Salt Lake City, Utah. Yao, S.S., Gilbert, J.R., 1984. Registration of a synthetic aperture radar image to thematic mapper imagery for remote sensing applications. IEEE Transactions on Geoscience and Remote Sensing GE-22 (6), 557 – 563. Zhang, Y., 1999. A new merging method and its spectral and spatial effects. International Journal of Remote Sensing 20 (10), 2003 – 2014.