Lab 5

Introduction

The goal of this lab was to introduce some analytical process: image mosaic, spatial and spectral image enhancement, band ratio, and binary change detection.  All these processes involve changing or converting an image to fit the needs of a study area.  Image mosaic is to teach how to process individual scenes of satellite images to arrive at one seamless scene.  At the end of the lab the goal was to be able to understand and use all of the analytical processes in real life situations or projects. 

Methods

In the methods sections I will explain the processes I took to arrive at an end result.  All process were done in ERDAS IMAGINE and one part in ArcMap.  Images and directions were provided by Dr. Cyril Wilson of the Geography Department and University of Wisconsin Eau-Claire. 

Part 1: Image Mosaicking

Opened Erdas Imagine and navigated to the image eau_claire_1995p26r29.   In the select layers to add multiple was clicked followed by multiple images in virtual mosaic, shown by figure 1 below.  

Figure 1: Multiple Window: Choosing Multple Images in Virtual Mosaic

Then also in the Select Layers to Add, Raster Options was chosen followed by choosing Background Transparent and Fit to Frame.  Then OK was hit to bring in the image.   The same exact process was done for the image eau_claire_1995p25r29 and the end result should be seen below in figure 2.

Figure 2: Both images brought in ready to mosaic them into one seamless tile. 

Next to mosoaic the images these steps were taken: Raster>Mosaic>Mosaic Express.  In the Mosaic image both images were brought in by clicking the folder in the middle of the page.  Eau_claire_1995p25r29 was brought in first because we want that image on top of the other.  Then next was clicked until the output tab was reached.  A new name was made and placed in a folder to be brought it.  Finished was clicked and the image was brought in to Erdas Imagine which you can see in the results tab below.

Section 2: Image mosaic with the use of MosaicPro

In this section a mosaic is going to be done with the same two images but this time an advanced mosaic tool called, MosaicPro was used to transform the image.

Again both images were added and adjusting the correct 'select layers to add' (see part 1) was done.  Then clicking on Mosaic>MosaicPro to bring up the MosaicPro window.  In this window the add images icon was clicked shown by arrow A below.  Next the image eau_claire_1995p25r29 was highlighted an the Image Area Options was edited.  Then Compute Active Area was clicked shown by arrow B below.

Figure 3: Image taken from Dr. Wilson

After hitting Ok the same steps were done to bring in the second image and two boxes appeared, see figure 3 above.   Next the select tool was hit, arrow B, to change the stack order of the images.  With the select tool click on the first image in, shown by arrow D, then send selected image to bottom, 3rd arrow C, to send the image to the bottom.  Repeat the process above to repeat the same process with the other image.  In the end the image eau_claire_1995p25r29 ended up on the bottom.

Figure 4; MosaicPro Image taken from Dr. Wilson

Next Color corrections was clicked, arrow E figure 4, to open the color corrections window.  Then Use Histogram Matching, arrow A figure 5, fill the box followed by set>and select overlap areas, arrow c figure 5 then OK to close it.  

Figure 5: Color Corrections and Histogram Matching, image
taken from Dr. Wilson.

To run the technique click process followed by Run Mosaic, the new image was named and saved in the correct folder and was brought into ERDAS IMAGINE.  The new image can be seen in the results section below.

Part 2: Band Rationing
In this part a band ratio will be performed by implementing the normalized difference vegetation index (NDVI) on eau_claire_2011 image.  That image was brought into ERDAS IMAGINE followed by raster>unsupervised>NDVI.  A new window appeared and filled out correctly, the sensor read 'Landsat TM' and under the select function NDVI was highlighted.  OK was clicked and the process was run.  The result can be seen below in the results section.

Part 3: Spatial and spectral Image enhancement
In part 3 different spatial enhancement techniques were used to transform images.

The image chicago_tm1995_b3 was brought in.  Under Raster>Spatial>Convulation a window was brought up like figure 6, below.

Figure 6: Convolution window.

The correct input and output file was set and the Kernel was set to 5x5 low pass.  The image was then compared side by side with the original image which can be seen below in the results. 

Next using the process and steps except with the sierra_leone2002 image and changing the kernel to 5x5 high pass.  The image again was brought into ERDAS IMAGINE to find results.  

This time using the edge enhancement tool the image sierra_leone1991 was used.  Using convolution again and 3x3 laplacian edge detection and checking Fill under handle edges by.  After clicking OK image was brought up to be compared to the old image.  

Section 2: Spectral enhancement

In this section linear contrast stretched was performed to improve the visual appearance of the images.  

The image eau_claire1991b3 was brought into and the histogram was analyzed.  Then pancromatic>General Contrast> General Contast to open up a new window.  Method was changed to Gaussain and apply was clicked to changed the image.  

Next, adding the second image eau_claire1991b5 was added.  Under Panchromatic>General Contrast>Piecewise Contrast was chosen.  The window and the range specifications were changed exactly to figure 7 below,

Figure 7: Piecewise Contrast Window
Image from Dr. Wilson 

This was then applied to the image and screenshot was captured, seen below in the results section.

Next a histogram equalization was performed on the image 15026029_0292011b30, the red band of Landsat TM of Eau Claire.   Raster>Radiometric>Histogram Equalization was clicked to open up a new window.  The input and output were set correctly and default values were accepted.  The function was run and a screenshot of the new image was captured to be seen in the results section.  

Part 4: Binary change detection

In part 4 calculations  were made to estimate the brightness values of pixels that changed from Eau Claire County from 1991 to 2011.  

2 images were brought in, ec_envs1991 and ec_envs2011.   Raster>Functions>Two Image functions was clicked to access two input operators window.  In input one image 1991 was entered and input 2 2011 was entered.  A new output name was created.  Under operator the + was changed to - and the layer file under both inputs was changed from All to Layer 4.   OK was clicked and the new image was screenshoted for results. 

Section 2: Mapping change pixels in difference image using spatial modeler.  

In this section a map was created that showed the differences between 1991 and 2011.  

The model maker toolbox was opened under Toolbox>Model Maker> Model Maker.

A model was created using the functions shown by figure 8 below. 

Figure 8: Arrows showing the different functions of model maker.

A= Selection tool, B= raster object C=function D=arrow E=where to place the items F=run the model

Figure 9: The model was set up like this, 2 rasters on top to
a function and then the output raster. 

ec_envs_2011_b4 was placed in the first raster and ec_envs_1991b4 was placed in the second.

In the function the first raster was subtracted by the second and 127 was added at the end.  A new output file was created for the final raster and the model was run.

 Then new image was then brought into ERDAS IMAGINE and calculations were made.  The standard deviation was times by 3 and added to the mean to give us the threshold number for a number of 202.18.  

Model maker was opened again with just two rasters and one function.  The new image that was just created was entered into the first raster.  In the function window analysis was changed to conditional and Either IF Or was chosen.   The equation- EITHER 1 IF ($n1_ec_91>202.18) OR 0 OTHERIWISE  as entered into the box.  An new output name was created and the image was opened.  

The two images, ec_envs_1991b4 and ec_91-11bvis were brought into ArcMap to give us a result, seen in the results tab below. 

Results

Part 1: Image Mosaicking

Figure 10: eau_claire1995msx

Describe the nature of color in your output image, in other words, is there a smooth color 

transition between one image and the other especially at the boundaries? 

1.      No, there is not a clean/ smooth color transition between the images.  The image on top is much darker in terms of red than the image below, it is easy to see the difference between the two.  

Section 2: Image mosaic with the use of MosaicPro

Figure 11: Eau Claire 1995msp

Compare the output mosaic image using the MosaicPro and that obtained earlier using the 

Mosaic Express. In your discussion, state the reason(s) for the differences in the image quality 

[hint: open two viewers and display each of the output images in a separate window] 

 The MosiacPro does a much better job of smoothing the colors together.  The edges of the pictures go together much smoother than the Mosaic Express technique.  If I were combing images I would use the MosaicPro technique.  This is because MosaicPro combines the two histograms to match the brightness of the images making it much harder to tell the difference in colors.  

Part 2: Band Rationing

Figure 12: Eau Claie 2000NDVI

What will you expect to find in areas that are very white in the NDVI image?

-Healthy Vegetation

Comment on the presence or absence of vegetation in areas that are medium gray and black. 

-It seems there is not much vegetation in areas around big cities, which makes sense since there are not very many agricultural fields near cites.   For example areas around the Twin Cities and Eau Claire are very dark compared to the rest of the image.  

Part 3: Spatial and spectral Image enhancement

Figure 13: Chicago 1995 5x5 Low

Outline the differences between the original image and the 5x5 Low Pass filtered image you 

just created.

-    The new image created is less bright, however, the new image has a lower spatial resolution and appears very blurry when zooming in.   Although, the high frequency of the image has improved making the image easier to interpret. 

Figure 14: Sierra 2002 5x5 High

Outline the differences between the original image and the 5x5 High Pass filtered image 

you just created. 

Although, the new image at first appeared very dark when zooming in there was a difference in brightness with the old image.  The old image is very gray all over, while the new image is black and white in places of rivers and city lights. 

Figure 15: Sierra 1991 Edge

Outline the differences between the original image and the Laplacian edge detection image 

you just created.

The old image is contrasted between to colors, red and green.  The old image appears not to use “real colors”.  The new image is much darker than the old and sort of blended the red and green together to make one color.  When zooming in the image cannot be interpreted at all making it not very useful because each pixel is a different color. 

Part 4: Binary change detection

Figure 16: EC 91_11

Section 2: Mapping change pixels in difference image using spatial modeler.  

Figure 17: Final product in Arc Map

Describe the spatial distribution of areas that changed over the 20 year period. Are these 

areas close to urban centers or not? 

The changes do not appear close to urban centers.  However, it depends on how you define ‘close’.  There does not appear to be any change close to Eau Claire, the biggest city in the area.  The changes to me seem to be in agricultural fields and close to a lot of water.  The changes do not have any pattern and are scattered throughout the area.