Lab 4 - Resolution and Contrast Enhancement

PART 1.

This part of the lab examines Landsat and SPOT imagery of San Diego to illustrate differences in spatial resolution and introduce filtering and contrast enhancement. Data files are found in \2002\lab5 in the class_lab folder on the desktop. Don't forget to email the lab answers to Jeff and Chris within a week.

In the main menu, select: File -> Open image file  and load Landsat_MSS_27Aug91  and Landsat_TM_year_1985  into two separate windows. Display the MSS image as RGB:2/3/1 and the TM as RGB:7/4/2. These are Landsat images of San Diego.

Q.1 Give examles of two types of feature that can be recognized in the TM but not in the MSS.

Now display the TM as RGB:7/4/1.

Q.2 Do you see a discernable difference in the image? Explain why or why not.

Make a scattergram of band 1 and band 2 (Tools ->2D scatter plots).

Q.3 Explain the scatterplot in the context of the differences you did or did not see when you changed bands.

Now do the same for MSS bands 1 and 2 (or 4 & 5 using conventional terminology).

Q.4 Explain the similarities and differences of this scatterplot compared to that of the TM.

Now apply a "low pass" (Filter -> Convolutions and Morphology-> Low Pass) to your TM data using the Apply to File button.  Use the defaults, output the result to memory and display the result in a new window using the same bands you are using for the original image.

Q.5 How does it look different?

Now apply a "high pass" (same sequence but High for Low) to the lowpass filtered image and display it as above in a new window.

Q.6 Does it look like the original you began with? How is it different?

Now, open a SPOT image of the same area:

In main menu select:
File -> Open image file  and load SPOT_Pan. Load "0.650_um":SPOT_Pan from Bands Available List and display as gray scale image. Increase the size of the image window. Image shows SPOT panchromatic data over San Diego.

Q.7 What is the pixel (ground resolution cell) size and what wavelengths are represented in SPOT panchromatic data?

From the Enhance menu examine the effectsof the various quick stretches on the tonal quality of the data in the image window as you pan around using the scroll window.

Q.8 What stretch (enhancement type) works best for you over most of the image area?

Q.9 What cultural features are resolvable in the SPOT_Pan data that were not recognisable in the Landsat TM image?
a) Street intersections?(Yes/No), b) Freeway interchanges?(Yes/No), c) Cars on freeway (Yes/No), d) Boats in water? (Yes/No), e) Waves breaking along beaches? (Yes/No), and f) Planes on airport runways? (Yes/No).

From the Enhance->  Interactive Stretching, and from Histogram_Source menu button, bring up the input and output histograms of the *SCROLL* window DN values. You'll notice that the input histogram is bimodal.

Q.10 What general cover types are associated with the two histogram 'modes'.

Experiment again with the different stretches as you pan around using the scroll window. Mask out the 'darker' mode value by moving the left dashed  line in the input histogram to the right to a DN value between 45 and 50. Re-do you favorite stretch and see if image quality over land areas is improved.

Q.11 Any improvement? (Yes/No/Maybe)

Using scroll window, position image window over waterway at west end of Coronado Island (big island with airport). Mask out the 'lighter' mode by moving the left dashed line left to a DN value of around 26, and move the right line left to about DN 47 and choose a linear stretch from the Sretch_Type menu.

Q.12 Describe what you see.

Q.13 Determine direction of travel of moving objects (use zoom window to advantage).

Q.14 Where is the bridge linking big island to mainland? (Hint: Pan around using scroll window to locate bridge).

Q.15 Did anyone pick out an artifact in the SPOT _Pan data? 

Please close all windows related to PART 1. The San Diego Tourist Board thanks you for your attention -- please come visit, the zoo is real good!

PART 2.

In this part we look at image data obtained from an airborne radar survey over the southern Alps of New Zealand, location of the Alpine Fault which separates the Australian and Pacific lithospheric plates.

In main menu select:

File -> Open External file -> image processing formats -> ER Mapper and load ts0431_c.dem_warped.ers. Enlarge the image window so that it reaches from the top to 2/3 the way to the bottom of your screen. This is a Digital Elevation Model (DEM) for topography obtained from radar interferometry along the flight line, which runs parallel to the long axis of the swath. The ground cell resolution (pixel) size is 10 m (same as SPOT Pan). From the functions menu in the image window select Color Mapping -> ENVI Color Tables, and change the gray scale to RAINBOW. In main menu select: Basic Tools->Cursor Location/Value, and read the data values within the black areas (areas within the swath where no backscatter returns were received), and the green areas OUTSIDE the borders of the swath (NULL pixels). Verify that these values are -16384 and -999, respectively. Close Cursor Location/Value window.

>From the Functions-> Display Enhancements select Interactive Stretching, and from Histogram_Source menu button, bring up the input and output histograms of the *SCROLL* window DN values. Look at the input histogram for the scroll window DNs. You'll notice (left-to-right) two spikes and a low, spread-out mode.

Q 15 Identify the three regions of the input histogram.

Change the lower limit of the input histogram to about 500 and the upper bound to about 12000, choose a linear stretch fron the Stretch_Type menu and press the `Apply' button.

Q.16 The `no backscatter' and NULL pixels are both now what color?

You can apply some of the other Stretch_Types to the DEM. Essentially, we've made a (pseudo)color map of topographic height.

In main menu select:

File -> Open External file -> image processing formats -> ER Mapper and load ts0431_c.vvi_warped.ers. This is vertically-polarised, C-band radar backscatter amplitude. Lighter shades of gray mean greater backscatter intensity.

Go to the Functions menu in the DEM image window select Links -> link Displays and link display #2 to display #1. The radar backscatter image window should resize to match the size of the DEM image window.

Q.17 What values do the `no backscatter' and NULL pixels take in the backscatter image?

In the radar backscatter window (display #2) choose Functions-> Display Enhancements select Interactive Stretching, and from Histogram_Source menu button, bring up the input and output histograms of the *IMAGE* window. Choose a value >0 (10, say) for the lower (red) bound of the input image histogram and examine the effects of various Stretch_Types as you pan around the swath using either of the scroll windows. (Note: Every time you pan, you need to reset the lower bound of the input image

histogram, and press `Apply').

Settle now on the Equalization Stretch_Type for the radr backscatter. Pan to the NE quadrant of the scroll window.

Q.18 What type of land use occurs in this wide, low area?

Q.19 What is the cover type in the hilly terrain of the middle and southern parts of the swath?

Radar is a *side-looking* active remote sensing technique, and the flight line runs parallel to the long axis of the swath. Using the dynamic overlay of the radar backscatter and DEM display windows determine

Q.20 From which side of the swath (right/left) is the terrain illuminated by the radar antenna?

[For you experts: Another way to decide this question is to directionally filter the topography and compare the resulting E->W or W->E gradient images with the backscatter data. From the main menu choose Filters-> Convolutions->Directional and apply to ts0431_c.dem_warped.ers in the data window. For W->E gradient (W-facing slopes

are brightest) choose Angle=0.0, for E->W gradients (E-facing slopes are brightest) chose Angle=180. Check OK, and store the filtered results In Memory as a new band. Load the new band in a new display window, link it to the other two and compare with the backscatter data (display #2). Truncate filtered data input histogram to lie between -256 and 256, and apply linear stretch. Good luck!]

Please close all windows related to PART 3. The `land of the long white

cloud' bids you good-bye. Stop by to see the kiwis next time you're in

the southern hemisphere.