Goal and Background
The goal of this lab was to become familiar with various ways of taking a raster image and changing/improving it. We learned:- Different ways to create an area of interest for our image
- Image fusion (pan sharpen)
- Simple radiometric enhancement techniques (haze reduction)
- Link the image to Google Earth
- Resampling
- Image mosaicking
- Binary change detection (image differencing)
Methods/Results (images)
We worked with the program ERDAS Imagine, for this lab. For each output image I created I made a subfolder within my lab 4 folder for better organization. To create our area of interest from the larger image I first had to go to the raster tools and then click inquire box. This created a nice box around the area I wanted to focus on, then I created a subset image from this. The only problem with this method is that most of the time the area of interest is not a perfect square/rectangle. For this part of the lab the eau_claire_2011.img image was used and the subset image created below is eau_claire_2011sb_ib.img.
Figure 1: Subset image from original eau_claire_2011.img image.
Next we created a subset image of the area of interest, this time using a shapefile. This created an area of interest that was fitted to the counties of Eau Claire and Chippewa Valley. For this section we had to add the Eau Claire and Chippewa Valley shapefile, which created a border on these counties in the eau_claire_2011.img. Again we went to the raster tools and clicked subset and chip to create the subset image. The images and shapefiles we used for this were, eau_claire_2011.img, ec_cpw_cts.shp, and the output image below is ec_cpw_2011sb_ai.img.
Next we created a subset image of the area of interest, this time using a shapefile. This created an area of interest that was fitted to the counties of Eau Claire and Chippewa Valley. For this section we had to add the Eau Claire and Chippewa Valley shapefile, which created a border on these counties in the eau_claire_2011.img. Again we went to the raster tools and clicked subset and chip to create the subset image. The images and shapefiles we used for this were, eau_claire_2011.img, ec_cpw_cts.shp, and the output image below is ec_cpw_2011sb_ai.img.
Figure 2: Subset image of Eau Claire and Chippewa County using shapefile.
Next we did something called image fusion (pan sharpen) which simply means we combined one image with the same image but in a higher resolution so the original image has color but is also clearer and less pixelated. To do this we went to the raster tools and then pan sharpen, and from there clicked on resolution merge from the drop down menu. We entered the appropriate parameters and created the image ec_cpw_2000ps.img. The high resolution image we used was ec_cpw_2000pan.img and the multispectral image we used was ec_cpw_2000.img.
Figure 3: Pan sharpened image (ec_cpw_2000ps.img)
Next we did some simple radiometric enhancement techniques. We reduced haze on an image that appeared foggy and unclear. We went to the raster processing tools and clicked haze reduction. The input file was eau_claire_2007.img and the output image we named ec_2007_haze_r.img. We accepted the default parameters.
Figure 4: Haze image (left) and Haze reduced image (right)
After this we learned how to link an image to Google Earth. This allowed us to pan and zoom into the same area on the image as well as Google Earth. This way we could compare images and identify objects in the image by using the higher resolution Google Earth image. To do this we opened the eau_claire_2011.img, clicked on the Google Earth tab and then clicked connect to Google Earth. We then synced the views so we could compare the Eau Claire image to the Eau Claire area in Google Earth.
Next we learned about resampling. Resampling is the process of changing the size of pixels. For this we clicked the raster tab, then spatial and from there resample pixel size. The input image was eau_claire_2011.img and the output image we named eau_claire_nn.img. This image was resampled using the technique nearest neighbor to change the pixel size from 30x30 to 15x15. We also tried the technique of Bilinear Interpolation this output image was called eau_claire_bli.img. Reducing the pixel size created a much clearer picture that did not pixelate as quickly as the original image when zooming in.
Image mosaicking was the next task we learned. For this we simply wanted to take two image and put them together in a way that made them look like they were one image stitched together. For this we used the images eau_claire_1995p26r29.img and eau_claire_1995p25r29.img. We mosaicked these two images together using mosaic express and mosaic pro. Both of these options are under the raster tools and then mosaic. Mosaic pro has many more parameters to enter than mosaic express which creates a nicer mosaicked image.
Figure 5: Image mosaicking using mosaic express (eau_claire1995msx.img)
Figure 6: Image mosaicking using mosaic pro, notice the images blend better using this technique. (eau_claire1995msp.img)
Lastly we learned about binary change detection (image differencing). We looked at the difference in brightness values of pixels over a 20 year period. To do this we clicked on functions and then two image functions. We used the images ec_envs2011.img and ec_envs1991.img, we named the output image ec_envrns_91_11.img. We used a short equation to figure out the difference in brightness of pixels between the 1991 image and the 2011 image. We used model maker to do the equation with the raster images. This created a simple model with input rasters, the equation and the output raster. After creating our image using the model maker we added it to the program ArcMap where we made the map slightly more appealing with a legend that clearly shows where brightness values changed and where they did not.
Figure 7: Histogram with lines drawn in showing the lower and upper limit of change/no change threshold.
Figure 8: Image showing change and no change of brightness values of pixels
