Geog 477-Lab2 Due September 28, 2006
Lab 2: Image Enhancement
In this lab you will learn to manipulate your data through image enhancement techniques for a more effective visual display and for easy interpretation of your image. For more information on these techniques, read: Jensen DIP ch. 5, 7, 8; Lillesand & Kiefer pg. 471-512.
Part I – Contrast Enhancement
Contrast enhancement effectively turns up the color contrast of an image by spreading the digital number (DN) values across the full range of possible numbers (0-255) by applying a function to the original values. Imagine can apply a contrast enhancement to your image through a variety of functions but in this lab we will look only at the Histogram Enhancement and Gaussian Stretch methods.
Original Image Contrast Enhanced Image
(from http://telsat.belspo.be/beo/en/guide/amcontr.asp?section=3.4)
(1) Histogram Enhancement
· Open a Viewer in Imagine. Go to File, select Open, and then select Raster Layer. On the File tab, select uncsubset.img from your directory, then click the Raster Options tab and check No Stretch. Then click OK to open the image in the Viewer. Now the image is displayed without stretch.
· Select Raster | Band Combinations from the Viewer menu to change the band combination to 3-2-1.
· Select Raster | Contrast | General Contrast from the Viewer menu.
· Click Help on the Contrast Adjust dialogue and read about the differences in various contrast enhancements.
· Select Histogram Equalization from the pull down menu. This enhancement allows you to assign an equal number of pixels to each of the output classes.
· Click on the Breakpts button. This allows you to see the frequency of pixels at each band. While the histograms appear as red, green or blue remember that they are displaying the band # that you assigned. So if in your band combinations you have NIR set to red, the red histogram is really showing NIR. If you have this in true color, the red histogram represents red, blue = blue, etc…. Make sure your Viewer, Breakpoint Editor and Contrast Adjust dialogue are all visible on your desktop.
· On the Contrast Adjust dialogue, hit Apply and observe what happens to the histograms. Now on the Breakpoint Editor dialogue hit Apply All and observe what happens to the image.
1. How are the ends of the histograms affected when you apply histogram equalization to the image? Why?
2. Which parts of the image (structures, forest, water, etc.) are most affected by the histogram equalization? Why?
3. Explain the histogram equalization method in your words.
· Close the Breakpoints Editor and Contrast Adjust dialogue. Do NOT undo the stretch and do NOT close the Viewer. Your will use this stretched image later on.
(2) Gaussian Stretch
· Open another Viewer. Open uncsubset.img in the new Viewer and make sure the image is NOT stretched.
· Select Raster | Band Combinations from the Viewer menu to change the band combination to 3-2-1.
· Select Raster | Contrast | General Contrast | Gaussian. Use the default setting with the Gaussian stretch.
· Open your Breakpoints Editor again. Click Apply on the Contrast Adjust dialogue and see what happens to your histogram.
· Hit Apply All on the Breakpoints Editor and see what happens to your image.
5. What happens to the ends versus the center of the histograms when you apply the stretch?
6. How is the readability of the image affected when you apply the stretch? Why?
7. In your own words, explain the Gaussian stretch method.
8. Compare the visual differences between the image stretched by the histogram equalization method and the image stretched by the Gaussian stretch method. Inspect Kenan Stadium, the pit, and other parts of campus. Which method would you prefer to represent the UNC campus? Why?
· Select Raster | Undo to undo the contrast you just applied to image.
Part II – Geometric Enhancement
Geometric Enhancement, also known as Filtering, can be used to improve the readability of an image. Filters use a Kernel to modify the DN value for each pixel according to the neighboring pixels’ values to make an image smoother, grainier, or to enhance certain spatial features of the image. In this lab, you will use a variety of Kernels to enhance your image geometrically.
Grainier Image Smoother Image
(from http://telsat.belspo.be/beo/en/guide/filtr.asp?section=3.5)
· Make sure that the uncsubset.img is open in a Viewer. Select Raster | Filtering | Convolution Filtering from the Viewer. The Convolve dialogue will appear on the screen. A number of filters are listed under Kernel.
· Apply each of the following filters to the image separately. To apply a filter, you select a kernel from the menu, and then click Apply. You can always undo what you did by selecting Raster | Undo. Examine the result when you apply each filter to the image.
§ 3*3 Edge Detect
§ 3*3 Edge Enhance
§ 3*3 Low Pass
§ 3*3 High Pass
§ Your Choice
9. Describe the visual effects of applying each of the filters above to the image, respectively. What spatial features does this filter enhance? What spatial features does this filter diminish? When would the use of this filter be appropriate?
· Create each of the following filters and apply each filter to the image. To create a new filter, click New on the Convolve dialogue. On the new dialogue, change the elements in the kernel. After you create a kernel, select File | Librarian. On the Kernel Librarian dialogue, specify the kernel name under Name: and save the kernel as a .klb file to your working directory. The new filter will be listed under Kernel: on the Convolve dialogue. Select the new filter and apply it to the image. You can undo what you did by selecting Raster | Undo.
Kernel 1 (Kernel Name: 3x3 hedge; File Name: hedge)
-1 / -1 / -10 / 0 / 0
1 / 1 / 1
Kernel 3 (Kernel Name: 3x3 diag; File Name: diag)
1 / 1 / 01 / 0 / -1
0 / -1 / -1
10. Describe the visual effects of applying each of the filters above to the image, respectively. How does this filter work? Why would you want to apply this filter to your image?
11. Create and name your own Kernel:
Describe a situation where this filter would be useful.
Part III: Paper Synopsis
Your task is to find a current paper that interests you from one of the following journals and write a brief synopsis (1 pg max) on the remote sensing research involved:
Remote Sensing of the Environment
International Journal of Remote Sensing
IEEE Transactions on Geoscience and Remote Sensing
The synopsis must describe the following elements: 1) the research questions; 2) types of methods used and a short description of each (field methods, statistical methods, etc.); 3) results found; and 4) the implications of the research.
************************************************************************
Your completed lab assignments should be saved to your directory as onyen_lab2.doc by 2 PM on the day that they are due. We will not accept paper labs.
Note that there are limits to available disk space. Please keep your directories clear of redundant and/or unnecessary files.
1