Image Band Transformations

Transcription

1 Image Band Transformations Content Band math Band ratios Vegetation Index Tasseled Cap Transform Principal Component Analysis (PCA) Decorrelation Stretch Image Band Transformation Purposes Image band transforms typically involve the manipulation of multiple bands of data, whether from a single multi-spectral image or from two or more images of the same area acquired at different times (i.e. multi-temporal image data). Image band transforms generate "new" images from two or more sources that highlight particular features or properties of interest, better than the original input images. Image Band Math You can perform mathematical operation on multi-spectral bands. In general, addition and multiplication operators tend to enhance correlated information between bands, while subtraction and division operators tend to enhance uncorrelated information between bands. Image subtraction is often used to identify changes that have occurred between images collected on different dates. Useful for mapping changes in urban development around cities and for identifying areas where deforestation is occurring. Band Ratios (Image Division) Image division or spectral band ratio is one of the most common mathematical operations applied to multi-spectral image data. Ratio images are calculated as the division of DN values in one spectral band by the corresponding pixel value in another band. Can be done in radiance, DN, emissivity, etc. Values can blow up (i.e., division by 0), so some scaling may be necessary to get best image quality Can take this further by rationing ratios Can combine ratio images to make RGB images Band ratio operation can reduce the environmentally induced variations in the DN values of a single band, such as brightness variations caused by topographic slope and aspect, shadows or seasonal changes in sunlight illumination angle and intensity. Therefore, band rationing tends to emphasize and highlight subtle variations in the actual spectral responses of various surface covers.

2 Enhances spectral differences and eliminates illumination differences, and takes advantage of spectral slopes. Image band ratio serves to highlight subtle variations in the spectral responses of various surface covers. By rationing the data from two different spectral bands, the resultant image enhances variations in the slopes of the spectral reflectance curves between the two different spectral ranges that may otherwise be masked by the pixel brightness variations in each of the bands. Band Rationing provide unique info not available in any single band that is useful for distinguishing soils and vegetation Band ratios for Landsat MSS For the four bands of the Landsat MSS, there are 12 different ratio combinations- 2/1,3/1,3/2,4/1,4/2,4/3, and their six reciprocals. The general utilities of MSS ratio are summarized as below. The 1/2,1/4,2/4, and 3/4 ratios are important for characterizing soil and rock units. Such an ordering would have vegetation depicted in dark tones; The 2/1 ratio is especially sensitive to the presence of iron oxide or ferric iron. In a 1/2 image, these same units would be depicted in dark tones; The 3/1, 3/2,4/1, and 4/2 ratios are useful for highlighting vegetation patterns because of the large differences in reflectance between the infrared bands (3 and 4) and visible bands (1 and 2); The 4/2 ratio is the most useful of the MSS ratios for assessing the relative greenness of vegetation, e.g., stressed plants versus unstressed and for estimating biomass. The 2/3 or 3/2 ratio is most often used for distinguishing general material types of soil and rock, vegetation, and water. Therefore, it is very useful for producing thematic maps of urban, soil and rock, vegetation, and water. Band Ratio for Landsat TM (ETM) For the six non-thermal bands of Landsat TM, there are 30 different ratio combinations-15 original and 15 reciprocal. General utility of several TM ratios are described as below. The 3/1 (red/blue) and 3/2 (red/green) ratios are important for delineating ferric iron-rich rocks (light-tones) and ferric iron-poor rocks (dark tones); Iron_Oxide ratio=(tm3-min(tm3))/(tm2- MIN(TM2)). The 5/7 ratio is useful for identifying clay-rich rocks (light-tones) because clay minerals exhibit strong absorption in the 2.2 µm region (band 7) and high reflectance in the 1.6µm region (band 5): Clay_ratio=(TM5-MIN(TM5))/(TM7- MIN(TM7)) The 4/3 ratio (near IR/red) uniquely defines the distinguishing different types of vegetation. Generally, the lighter the tone, the greater the amount of vegetation present. The 5/2 ratio (mid IR/green) is useful for distinguishing different types of vegetation. Its reciprocal is useful for identifying water bodies and wetlands. The 3/7 ratio (red/mid IR) is useful for observing differences in water turbidity. It is also useful for identifying the roads and other cultural features, appearing

3 lighter tone due to their relatively high reflectance in the red band (TM3) and low reflectance in the mid IR band (TM7). Vegetation Index Various forms of ratio combinations in the wavelength range µm (near IR) to those in the µm range (red) have been developed for vegetation monitoring, e.g., assessing biomass or leaf area index and discriminating between stressed and non-stressed vegetation. Distribution of Pixels in a Scene in Red and Near-infrared Multispectral Feature Space Reflectance Curves for Selected land cover types Simple vegetation index (VI) sensitive indicators of the presence and condition of green vegetation: VI = NearIR-Red Infrared/Red Ratio It takes advantage of the inverse relationship between chlorophyll absorption of red radiant energy and increased reflectance of near-infrared energy for healthy plant canopies. Normalized Difference Vegetation Index (NDVI) the most commonly used index provided a method of estimating net primary production over varying biome types (e.g. Lenney et al., 1996), identifying ecoregions (Ramsey et al., 1995), monitoring phenological patterns of the earth s vegetative surface, and of assessing the length of the growing season and dry-down periods (Huete and Liu, 1994). This index largely compensates for changing illumination conditions, surface slope, and viewing aspects. Healthy vegetation reflects strongly in the near-infrared portion of the spectrum while absorbing strongly in the visible red. Other surface types, such as soil and water, show near equal reflectances in both the near-infrared and red portions. Thus, the NDVI image would significantly enhance the discrimination of vegetation from other surface cover types. Also, we may be better able to identify areas of unhealthy or stressed vegetation, as the ratios would be lower than for healthy green vegetation. In general, vegetation yields high positive NDVI values. Clouds, water, and snow yield negative values due to larger red reflectance than near IR. The NDVI values for rock and bare soil are near zero due to their similar reflectance in both bands. Therefore, in a NDVI image the lighter tones are associated with dense coverage of healthy vegetation. The red and near infrared bands that are commonly used for the NDVI calculation are listed below: AVHRR: band 1 for red, and band 2 for near IR. Landsat MSS: band 2 for red, band 4 or band 3 for near IR; Landsat TM: band 3 for red, band 4 for near IR;

4 SPOT HRV: band 2 for red, band 3 for near IR; Infrared DOQQ: band 2 for red, band 3 for near IR Infrared Index An Infrared Index (II) that incorporates both near and middle-infrared bands is sensitive to changes in plant biomass and water stress in smooth cord grass studies (Hardisky et al., 1983; 1986): Healthy, mono-specific stands of tidal wetland such as Spartina often exhibit much lower reflectance in the visible (blue, green, and red) wavelengths than typical terrestrial vegetation due to the saturated tidal flat understory. In effect, the moist soil absorbs almost all energy incidents to it. This is why wetland often appears surprisingly dark on traditional infrared color composites. Soil Adjusted Vegetation Index (SAVI) Recent emphasis has been given to the development of improved vegetation indices that may take advantage of calibrated hyperspectral sensor systems such as the moderate resolution imaging spectrometer - MODIS (Running et al., 1994). The improved indices incorporate a soil adjustment factor and/or a blue band for atmospheric normalization. The soil adjusted vegetation index (SAVI) introduces a soil calibration factor, L, to the NDVI equation to minimize soil background influences resulting from first order soil plant spectral interactions (Huete et al., 1994): An L value of 0.5 minimizes soil brightness variations and eliminates the need for additional calibration for different soils (Huete and Liu,1994). Soil and Atmospherically Adjusted Vegetation Index (SARVI) Huete and Liu (1994) integrated the L function from SAVI and a blue-band normalization to derive a soil and atmospherically resistant vegetation index (SARVI) that corrects for both soil and atmospheric noise. The technique requires prior correction for molecular scattering and ozone absorption of the blue, red, and near-infrared remote sensor data, hence the term p*. Enhanced Vegetation Index (EVI) The MODIS Land Discipline Group proposed the Enhanced Vegetation Index (EVI) for use with MODIS Data: The EVI is a modified NDVI with a soil adjustment factor, L, and two coefficients, C1 and C2 which describe the use of the blue band in correction of the red band for atmospheric aerosol scattering. The coefficients, C1, C2, and L, are empirically determined as 6.0, 7.5, and 1.0, respectively. This algorithm has improved sensitivity to high biomass regions and improved vegetation monitoring through a decoupling of the canopy background signal and a reduction in atmospheric influences (Huete and Justice, 1999). Moisture Vegetation Index Rock et al (199) utilized a Moisture Stress Index (MSI) based on the Landsat TM near-infrared and middle-infrared bands Tasseled Cap Transformation Kauth-Thomas Tasseled Cap transformation is based on Gram-Schmidt sequential orthogonalization techniques that produce an orthogonal

5 transformation of original four-channel MSS data space to a new fourdimensional space. It is called the tasseled cap transformation due to its cap shape. Kauth-Thomas Tasseled Cap transformation For Landsat MSS After the Tasseled Cap transformation of Landsat MSS data, four new separate images are created, including the soil brightness index (SBI), the green vegetation index (GVI), the yellow stuff index (YVI), and a non-such index (NSI) associated with atmospheric effects and image noise. Generally, the GVI and SBI indexes contain most of the scene information (95% to 98%). The Tasseled Cap (TC) transformation provides excellent information for agricultural applications because it allows the separation of vegetated surface from barren (bright) and wet soils. The Tasseled Cap transformation for the Landsat MSS data can be performed using the following formula: SBI=0.332MSS MSS MSS MSS4 GVI=-0.283MSS MSS MSS MSS4 YVI=-0.899MSS MSS MSS MSS4 NSI=-0.016MSS MSS MSS MSS4 For Landsat TM Tasseled Cap transformation has also extended to Landsat TM data. Four separate images can be created based on the application of Tasseled Cap transformation on six non-thermal bands of Landsat TM data: scene brightness, vegetation greenness, surface wetness, and atmospheric haze. The Tasseled Cap transformation of Landsat TM data can be performed using the following formula: Brightness= TM TM TM TM TM TM7 Greenness= TM TM TM TM TM TM7 Wetness = TM TM TM TM TM TM7 Haze = TM TM TM TM TM TM7 Principal Component Analysis (PCA) Individual bands of a multi-spectral image or images from multi-sensors over the same area are often highly correlated. They appear similar and convey essentially the same information. Principal component analysis (PCA) is most commonly used techniques to remove or reduce redundancy in multi-spectral data. PCA is a linear transformation that produces a new set of bands, called components, through a weighted linear combination of the original image bands. For any given number of original bands, an equivalent number of component images can be created by PCA. The resulting PC images are uncorrelated with one another and ordered in terms of their explanatory power (variance). Each subsequent PC accounts for an increasingly small amount of variation in original data. Transformation of the raw remote sensor data using PCA can result in new principal component images that may be more interpretable than the original data. Since the first few principal components often contain over 90 percent of the information in the original bands, the remaining components can be dropped in

6 the subsequent analysis without significant loss of the information. In this way, PCA reduces the dimensionality (the number of bands) of the original data, and compress information (or variance) from the original data into the least number of new components. Principal components analysis, and other complex transforms, can be used either as an enhancement technique to reduce the number of bands to be used as input to digital classification procedures or to improve visual interpretation. Canonical Analysis CA uses the spectral characteristics of categories within the data to increase their separability. CA uses the spectral characteristics of categories within the data to increase their separability. Combine PC bands as RGB; PCA band 1 is commonly dominated by temperature in IR, brightness/shadowing in VNIR, both result from topography To enhance variation in PC images, apply stretch to PC bands, rotate back to original axis and display as image Suggested reading Chapter 8 in Jensen, J.R Introductory digital image processing: a remote sensing perspective, 3rd Edition. Upper Saddle River, NJ, Prentice Hall. 526 pp Please see the class slides for details