IMAGINE LAB 1. Introducing ERDAS IMAGINE, Spectral Properties of Landsat Images, and Image Enhancement

Imagine Lab-1: Pg. 1 ENVIRON/NRE441 (W06) 02/13/06 IMAGINE LAB 1. Introducing ERDAS IMAGINE, Spectral Properties of Landsat Images, and Image Enhancement This week you are going to learn several functions of the industry standard image processing software ERDAS IMAGINE. These will help you to explore Landsat-7 ETM+ images of Washtenaw County. The images, which cover the Pinckney and Stinchfield Woods Area, were taken on July 15, 1999. Specifically you will learn: 1. Basics of using the ERDAS IMAGINE software 2. Several standard image composites using Landsat ETM+ (Enhanced Thematic Mapper Plus) 3. Spectral properties of several land-cover types and their spectral reflectance curves as sensed by Landsat ETM+ imagery 4. How to evaluate remote sensing image and band characteristics using histograms 5. How to enhance remote sensing images (to increase contrast and dynamic range) 6. How to compute a normalized difference vegetation index (NDVI) from Landsat ETM+ Preparation: You should have read L&K 491-517, 531-545; and Jensen pp. 140-152, 179-186. The output for this Lab: 1. Your answers to questions asked in this lab exercise 2. One graphic that you copy and paste into your report (a spectral curve) 3. One table that you fill out from this exercise Hint: You may want to start and open a MS Word document while you are doing this Lab into which you can type answers and paste the screen shot graphics. There is a MS Word template on the CTools website that has the questions and table for the lab write-up. Hint: Read ahead the Questions to keep them in mind while doing the exercise. Setup your lab directory: 1. In Windows Explorer create a folder called ?imagine1? in H:\nre441, if you haven?t done so. Download lab data and write-up sheet from the CTools web site. 2. When you need to save any images or data you will also save them to H:/nre441/imagine1 folder. You should always check the directory whenever you try to save an ERDAS IMAGINE file. 3. File naming convention: always precede any files you make with your initials followed by an underscore and then the suggested filename (e.g. tz_stinchfield.img or kb_stinchfield.img). This helps us to identify your work. Imagine Lab-1: Pg. 2 I. USING THE ERDAS IMAGINE VIEWER The IMAGINE viewer is where most of your interaction with remote sensing imagery in IMAGINE will take place. To start we briefly learn its basic functions and how to load and examine a Landsat image. STEP 1: Loading Imagery ? In Viewer #1, from the Viewer menu, click File | Open, then select Raster Layer. . . The Select Layer to Add dialog box displays. ? Verify that File Type is set to *.img. Click Goto. ? Under Select A Directory dialog, highlight afs/?/[your uniquename]). Click OK. ? In the Directory section of Select Layer to Add dialog, navigate to the nre441/imagine1 folder. ? Highlight etmp.img. ? Click the Raster Options tab to examine the display parameters. Confirm that the Display as is set to True Color. Change Layers to Colors to 4, 3, 2 if not already set. Note that 4 = TM 4 (NIR) 3 = TM 3 (Red) 2 = TM 2 (Green) Imagine Lab-1: Pg. 3 ? Click OK and the image displays in the Viewer ? Right click in the Viewer #1 to display the QuickView options, and then click on Fit Image to Window STEP 2. Roaming and zooming ? In the Viewer menu, select View | Zoom | In by 2 ? From the Viewer tool bar, click the Roam icon. Notice that the pointer, when placed in the Viewer, changes to hand. ? Left click, drag, and roam around the image. STEP 3. Setting the Viewer Preferences ? We can set the Raster Option dialog box (as showed in the above figure) so that it displays with the Fit to Frame option set as the default. ? In the IMAGINE menu, click Session | Preferences. The Preference Editor displays. ? In the Category list, select Viewer, and the preferences for the Viewer display in the right section of the same window. ? Scroll down to the Fit to Frame option. As a default, the option is disabled. Enable the checkbox to set the new default for the Raster Option dialog box. ? Click the User Save button, and then click Close. Imagine Lab-1: Pg. 4 STEP 4. Viewing Image properties ? Click on Viewer from the IMAGINE Icon Panel to open another viewer ? Select File | Open | Raster Layer. . . ? Open the Landsat image etmp.img in this new Viewer ? Pull Down the Utility manu and select Layer Info ? Examine the ImageInfo screens, including each of the four Tabs. We will investigate these further in the remaining part of this Lab. ? General ? Projection ? Histogram ? Pixel data STEP 5. Close ImageInfo by clicking File | Close. Close Viewers by clicking File | Close of each open Viewer. Imagine Lab-1: Pg. 5 II. Landsat ETM+ Data and Digital Data Manipulation You?ve been introduced in very brief fashion to the IMAGINE viewers. The next lab steps will give you a more in-depth explanation of these and several basic IMAGINE functions related to Landsat data and its spectral properties. 1. Landsat-7 ETM+ (Enhanced Thematic Mapper ? Plus) Landsat-7 was launched on April 15, 1999 from Vandenberg, AFB in California. Onboard Landsat-7 is the ETM+ sensor. The table below summarizes the spectral and spatial properties of ETM+ images. Table 1. Band number Band description Bandwidth (?m) Pixel (meter) 1 Blue 0.45-0.52 30 2 Green 0.52-0.60 30 3 Red 0.63-0.69 30 4 Near infrared 0.76-0.90 30 5 Mid-infrared 1.55-1.75 30 6 Thermal infrared 10.4-12.5 60 7 Mid-infrared 2.08-2.35 30 Panchromatic Green-red-NIR 0.52-0.90 15 The radiometric resolution for ETM+ bands is 8-bit, that is, the values of digital numbers (DNs) in each band range from 0 to 255. Landsat-7 has a 16-day repeat time to take images of the same location on earth. Different bands have different radiometric responses to the biophysical characteristics (in our case land-cover) of earth surface. This exercise will help you to understand the relation between the DNs and land-cover characteristics. 2. Viewing the image using color composites In an IMAGINE Viewer, open the etmp.img with the true color band combination, i.e., display image band 1 (blue band) using monitor blue color, band 2 (green band) using green color, and band 3 (red band) using red color. To do this, first, click on the Viewer icon on the IMAGINE icon panel. Select the File | Open | Raster Layer to bring up the Select Layer to Add dialog. Navigate to your directory and highlight etmp.img. Click on the Raster Options tab in the dialog. You will see the default settings of the display band combination to be Red: 4, Green: 3, and Blue: 2. Change Red color assignment to 3, Green to 2, and Blue to 1 and click OK. Imagine Lab-1: Pg. 6 The image you are seeing in the Viewer should be similar to what one would see with naked eyes. The land-cover features appear as follows: Table 2. Land-cover Display color Trees, bushes Olive green Crops, grass Medium to light green Wetland Dark green to black Water Shades of blue and green Urban areas White to light blue Bare soil White to light blue Take a look at examples of each of these. Please also locate the business area of Pinckney (close to the upper left edge of the image) and Stinchfield Woods (roughly at the center of the image). Now open another viewer (viewer #2) and load the same etmp.img with default band combination, i.e., Red: 4, Green: 3, Blue: 2. The 432 band combination is called a Near Infrared Color Composite or Standard False Color Composite. It?s one of the most commonly used combinations. Land-cover features appear in a near infrared color composite as follows: Table 3. Land-cover Display color Sample locations on etmp.img (X: column, Y: row) Trees, bushes Red (134,254), (299,157) Crops, grass Pink to red (324,367), (310,199) Wetland Dark red (25,116), (228,211) Water Shades of blue or black (76,189), (190,185) Urban areas Cyan to gray (126,37), (208,26) Bare soil Cyan to gray (73,324), (251,278) The following steps will help you to interpret land-cover types at different locations. First, select View | Link/Unlink Viewers | Geographical from the menu of viewer #1. The Link/Unlink Instructions message window appears. Now, move the cursor to viewer #2. The cursor will change its shape to . Just click within the viewer #2 window to link both viewers. Right click in viewer #2 and select the Inquire Cursor? to display the cursor inquiry window (you may move windows around on the desktop to make them more convenient to see). This window displays information about the actual DN value of each pixel in the image file (FILE PIXEL), the DN value displayed on screen (LUT VALUE), and the total number of pixels with Imagine Lab-1: Pg. 7 that DN number in the file (HISTOGRAM). A white cursor (cross) also appears in both viewers. The location of the cursor is shown in the X and Y fields in the cursor inquiry window. With this inquiry tool, we can find the locations of land-cover features listed in Table 3. First, change the Map setting in the selection list on the upper left corner of the cursor inquiry window to File. Type in the X and Y (column and row) values listed in Table 3, e.g., 134 and 254 is a tree/forest land-cover type. See how trees look in a near infrared color composite and a true color image. Go through all land-cover types listed in Table 3. Zoom in and out if you need to. Now, you should be able to interpret different land-cover types by just looking at the near infrared color composite. Question #1: Please comment on the advantages and disadvantages of using these two different band combinations to interpret land-cover types. Imagine Lab-1: Pg. 8 Another commonly used band combination is 742 (Red: 7, Green: 4, Blue: 2). It differentiates land-cover features as well as a near infrared color composite does and displays vegetation as green, which looks more realistic than the near infrared color composite. To change the band combination of a viewer after the image is already loaded (e.g., viewer #1), select Raster | Band Combinations? from the menu of the Viewer and set the intended band combination, e.g., Red: 6 (the 7 th band in ETM+), Green: 4, and Blue: 2. Because the thermal infrared band of original ETM+ has a different spatial resolution, it hasn?t been added to our etmp.img file. The 7 th band (MIR2) is the 6 th band in the file. Click Apply then Close. Figure 1 below is a typical spectral reflectance curves for vegetation, soil, and water. You will use the spectral profile tool to create a similar diagram and compare it with Figure 1. Figure 1. (Source: Swain, P.H. and Davis, S.M. (eds) 1978. Remote Sensing: The Quantitative Approach. MaGraw- Hill, NY.) Imagine Lab-1: Pg. 9 Select Raster | Profile tools? from the menu of etmp.img viewer #1. Select Spectral option and click OK. After the SPECTRAL PROFILE #1 window appears, click on the Create New Profile Point in Viewer icon . Move your cursor to the viewer #1 window and click on a location that is tree cover (i.e., forested). Repeat the creation of profile points for soil and water land-cover types. Hint: it may be convenient to follow the order of land cover types in Table 3. You will need to edit the legend on the spectral profile later. Hint: To find the representative pixels for different land-cover types, you may want to use those locations listed in Table 3. An easy way to locate those positions is to enter point coordinates in the blank number fields on the Spectral Profile window. Here to put X and Y values from Table 3 Imagine Lab-1: Pg. 10 After creating profile points for three land-cover types, you could select Edit | Chart Legend? to edit the name (note: you?ll need to hit Enter after you input the new name here to make it effective) and color of the legend of your profile diagram. If you need to delete any profile, use Edit | Delete Plots, and then click to place an ?X? in the Delete Plot column of the one you wish to delete and click on OK to delete it. Question #2: Please compare and comment on the similarities and differences between your diagram and Figure 1. Also incorporate a copy of your diagram in your write-up. Hint: How to put images (graphics) into your report? Select the window that you want to use in your report. Press the ?alt? key and hit ?Print Screen/SysRq? key while the ?alt? key is still pressed. This ?captures? the image of the active window on your screen. Now, open your report with a word processor (e.g., MS Word), then move the cursor to the location you want to insert the captured image and press ?ctrl-V? (or select ?Paste? from the edit menu). The image will be inserted to your document. Save your file. 3. Histograms of a digital remotely sensed image Histograms are another useful statistical tool to describe the image characteristics. The X (horizontal) axis of a histogram represents a range of values in data. These are usually the DNs of the actual file values and/or screen values. The Y (vertical) axis of a histogram shows the number of pixels at each data value, which is called the frequency of each value. By examining the histogram of an image band, we can have a general idea of the distribution of the DNs of that band. This in turn is an indicator of what kinds of land-cover and how much are they in the Imagine Lab-1: Pg. 11 image area. You will use the Imageinfo dialog to complete the following table and interpret the meaning of this table. Table 4. Band mean min max range (max-min) Std Dev. Histogram 1 (Blue) 2 (Green) 3 (Red) 4 (NIR) 5 (MIR1) 6 (MIR2) Please select Utility | Layer Info? from the menu of viewer #2 (your standard false color infrared composite) to open the Imageinfo dialog. From the menu of the Imageinfo dialog, select Edit | Compute Statistics?, and set X and Y skip factors to 1, then click OK. This will take all pixels in the image into account when calculating the descriptive statistics. Browse through all bands in the Imageinfo dialog using the navigation arrows and fill in the blanks in Table 4. Note you can get information on all the six bands even though you are only seeing three of them displayed in the Viewer. At the same time, you can select View | Histogram?, or click on the Display the Layer Histogram icon , or click on the Histogram tab in the Imageinfo dialog to display the histogram of the band selected. Imagine Lab-1: Pg. 12 When the pointer (controlled by the mouse) is in a histogram, the numbers that appear near the graph show you values that are represented by the pointer location. At bottom center, the data value represented by the pointer location is shown. At left center, the frequency of that data value is shown. The min, max, and mean DN values are also showed in the histogram window. Make a quick sketch of the histograms (or take a screen shot and paste it) in the histogram column of Table 4. Question #3: Please describe the distributions of the six ETM+ bands, e.g, which one(s) got the largest range, which one(s) got lower DN values. Explain why these histograms are different by associating Figure 1 with Table 4. How does the atmospheric scatter effect influence the distribution of these histograms? What land-cover does the spike on the left side of the IR (NIR and mid-IR) band histograms represent? 4. Contrast enhancement After an image is displayed in a viewer, several methods are available to adjust the contrast of the display to enhance certain features of the image. The procedure of adjusting the contrast of an image is referred as contrast enhancement (also called a stretch). By default, ERDAS IMAGINE displays an image with a standard stretch. Let?s look at how an image appears without that stretch. Close all open viewers and bring up a new one. Click on File | Open | Raster Layer in viewer #1 to bring up the Select Layer to Add dialog. Highlight etmp.img. Bring up the Raster Options tab and enable the No Stretch checkbox, then click OK to accept the default band combination with no stretch. Imagine Lab-1: Pg. 13 Usually the original image has less contrast than the maximum contrast a display device allows. It?s because remote sensing sensors must be sensitive to reflectance from diverse biophysical materials such as dark volcanic basalt outcrops or bright white snow. However, very few scenes are composed of DN values that utilize the full sensitivity range of the sensors. Contrast enhancements allow us to expand the range of original limited brightness values and use the full range of the display lookup table (LUT). Your viewer does a default stretch automatically unless you specify no stretch as you just did. Before doing the contrast enhancements, we need to set up a display environment that allows you to visually compare the enhanced and the original images. In viewer #1 open etmp.img (with stretch method) as a second layer. To do that, click on File | Open | Raster layer, in the Raster Options tab disable the Clear Display check-box. Now select View | Arrange Layers from the menu of the viewer to open the Arrange Layers dialog. With this dialog, you can control the display of the layers listed in it. In viewer #1, you should see the stretched image (i.e., brighter image) now. The image listed on the top of the list in Arrange Layers dialog is this active and visible layer. To change the active layer to the no- stretch layer, put the mouse pointer on the second layer in the list, then right click and select Raise To Top. Click on the Apply button to make the change effective. Since we need to keep the original image as a reference, change the active (visible) layer back to the stretched image with the same procedures. Using the stretched image, now select the Raster | Contrast | General Contrast? from the menu of viewer #1. The Contrast Adjust dialog appears. Click on the Breakpts? button to bring up the Breakpoint Editor dialog. In the Breakpoint Editor dialog, the original and stretched histograms are displayed using gray and color legends, respectively. The dash line in each histogram window is the lookup table graph that maps the relationship between the DN values in the file (horizontal axis) and the values display on screen (vertical axis ranging from 0-255). The other number on the vertical Imagine Lab-1: Pg. 14 axis is the frequency of histogram. For example, the diagram below shows the original DN values, ranging from 47 to 87, have been mapped to the full range of display (0 to 255). All DN values smaller than 47 are displayed as 0 (black) and those larger than 87 are displayed as 255 (blue). The reason to do such contrast stretch is that there are not many pixels in the file with DN values smaller than 47 or larger than 87, as is shown by the histogram. There are two major groups of contrast enhancement methods: linear stretch and nonlinear enhancement. The Standard Deviations method available from the Contrast Adjust dialog performs a standard deviation linear stretch on the range of the lookup table. The Histogram Equalization method is a nonlinear stretch that redistributes pixel values so that there are approximately the same number of pixels with each value within a range. The result approximates a flat histogram. Therefore, contrast is increased at the ?peaks? of the histogram, and lessened at the ?tails?. Histogram equalization can also separate pixels into distinct groups, if there are few output values over a wide range. This can have the visual effect of a crude classification. Now, let?s see how the contrast enhancement changes the look of the image. Select Standard Deviations from the Contrast Adjust dialog, leave the default of standard deviations at 2.0, and click on the Apply button. Notice how the display histograms in the Breakpoint Editor changes. Click on the Apply All button on the Breakpoint Editor dialog and see how the image in viewer #1 changes. By reading the lookup table graph careful, you know what display DN value is used for a given DN value in the file. ERDAS IMAGINE provides several utilities to facilitate visual comparison of images. We will use the Flicker and the Swipe tools to compare the enhanced and original images. Go to viewer Imagine Lab-1: Pg. 15 #1 and select Utility | Flicker. Click on the Manual Flicker button and see what happens to the viewer. Click the button again and see what happens. Check the Auto Mode option and see what happens. To use the swipe tool, click Cancel on the Viewer Flicker dialog and select Utility | Swipe from the menu of viewer #1. Experiment on the options and controls of the dialog by yourself. When done, click Cancel button to close the swipe utility. Next we will use Histogram Equalization method to enhance the image. Before you do anything, select Raster | Undo from the menu of viewer #1 (this undos the last contrast enhancement). Then, repeat the same procedures described above using the Histogram Equalization enhancement method rather than Standard Deviation. Compare the enhanced and original images with flicker or swipe utilities. Question #4: Explain the difference between linear stretch and histogram equalization. Close the Breakpoint editor, Contrast Adjust and Arrange Layers windows. Close the image by File | Close. DO NOT save changes!!! Note: we do not want the permanent contrast enhancement. Permanent contrast enhancement refers to the adjustment of the DN values in image files and overwriting the original values in the actual data file. On most occasions (including our lab), contrast enhancement is only for display purposes. Enhanced DN values are not suitable for analyses. You have now finished your first ERDAS IMAGINE lab. Congratulations! It is always important to exit IMAGINE properly by using the Session | Exit IMAGINE command before you shut down your computer. This will ensure the IMAGINE license that you?ve been using not to ?hang-up? (if it is hanging up there, you might not be able to run the software the next time due to the limited license sits). You also want to exit (by typing ?exit? and hitting Enter) thor and cmd windows. tzhao Microsoft Word - nre441imaginelab1_06.doc