mypapers/drafts/geg2210/assignment-8.html

   1 <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
   2 <html>
   3   <head>
   4     <title>Assigment 8 in GEG2210 2005</title>
   5   </head>
   6
   7   <body>
   8
   9     <p><a href="http://www.geo.uio.no/geogr/geomatikk/oppgaver/bildeforbedring_eng.html">Assigment 8</a>
  10       in <a href="http://www.uio.no/studier/emner/matnat/geofag/GEG2210/index-eng.html">GEG2210</a>
  11       - Data Collection - Land Surveying, Remote Sensing and Digital
  12         Photogrammetry</p>
  13
  14     <h1>Image enhancement, filtering and sharpening</h1>
  15
  16     <p>By Petter Reinholdtsen and Shanette Dallyn, 2005-05-01.</p>
  17
  18 <p>This exercise was performed by logging into jern.uio.no using ssh
  19 and running ERDAS Imagine.  Started by using 'imagine' on the command
  20 line.  The images were loaded from /mn/geofag/gggruppe-data/geomatikk/</p>
  21
  22 <p>We tried to use svalbard/tm87.img, but it only have 5 bands.  We
  23 decided to switch, and next tried jotunheimen/tm.img, which had 7
  24 bands.</p>
  25
  26 <h2>Some notes on the digital images</h2>
  27
  28 <p>The pixel values in a given band is only a using a given range of
  29 values.  This is because sensor data in a single image rarely extend
  30 over the entire range of possible values.</p>
  31
  32 <p>The peak values of the histograms represent the the spectral
  33 sensitivity values that occure the most often with in the image band
  34 being analysed.</p>
  35
  36 <h2>Evaluation of the different bands</h2>
  37
  38 <p><img src="jotunheimen-truecolor.jpeg" align="right" width="40%">
  39 This image show the "true colour" version, with the blue range
  40 assigned to the blue colour, green range to green colour and red range
  41 to red colour.</p>
  42
  43 <h3>band 1, blue (0.45-0.52 micrometer - um)</h3>
  44
  45   Visible light, and will display a broad range of values both over
  46   land and water.  Reflected from ice, as those are visible white and
  47   reflect all visible light waves.  Histogram show most values between
  48   30 and 136.  Mean values of 66.0668.  There are one wide peak with
  49   center around 50.  There are two peaks at 0 and 255.
  50
  51 <h3>band 2, green (0.52-0.60 um)</h3>
  52
  53   Visible light, and will display a broad range of values both over
  54   land and water.  Reflected from ice, as those are visible white and
  55   reflect all visible light waves.  Histogram show most values from 8
  56   to 120.  The mean value is 30.9774.  There are two main peaks at 20
  57   and 27.  There is also a pie at 0.
  58
  59 <h3>band 3, red (0.60-0.69 um)</h3>
  60
  61   Visible light, and will display a broad range of values both over
  62   land and water.  Reflected from ice, as those are visible white and
  63   reflect all visible light waves.  Histogram show most values from 33
  64   t 135, with one wide peak around 52.  There are also seem to be two
  65   peaks at 0 and 255.  The mean value is 34.3403.
  66
  67 <h3>band 4, near-infraread (0.76-0.90 um)</h3>
  68 <img src="jotunheimen-band4-hist.jpeg" align="right" width="20%">
  69
  70   Water acts as an absorbing body so in the near infrared spectrum,
  71   water features will appear dark or black meaning that all near
  72   infrared bands are absorbed. On the other hand, land features
  73   including ice, act as reflector bodies in this band.  The histogram
  74   show most values between 7 and 110.  The mean is 40.1144.  There are
  75   two peaks at 7 and 40.
  76
  77 <h3>band 5, mid-infrared (1.55-1.75 um)</h3>
  78
  79   The ice, glaciers and water do not reflect any mid-infrared light.
  80   The histogram show most values between 1 and 178.  The mean is
  81   49.8098 and there are two peaks at 6 and 78, in addition to two
  82   peaks at 0 and 255.
  83
  84 <h3>band 6, thermal infrared (10.4-12.5 um)</h3>
  85
  86   Display the temperature on earth.  We can for example see that the
  87   ice is colder than the surrounding areas.  The histogram show most
  88   values between 36 to 122.  The mean is 102.734.  There are one wide
  89   peak around 53, in addition to two peaks at 0 and 255.
  90
  91 <h3>band 7, mid-infrared (2.08-2.35 um)</h3>
  92
  93   The ice, glaciers and water do not reflect any mid-infrared
  94   frequencies.  The histogram show most values between 77 and 150.
  95   The mean is 24.04, and there are one wide peak at 130 and a smaller
  96   peak at 83, in addition to one peak at 0.
  97
  98 <h3>Image enhancement</h3>
  99
 100 <img src="jotunheimen-std-ir-lin.jpg" width="40%">
 101 <p>When we look at the linear contrast functions, we can move the
 102 slope and shift values increasing or decreasing the contrast of the
 103 image. For example, in the linear contrasting we moved the slope value
 104 from 1.00 to 3.00 to obtain a brighter appearing image, and then we
 105 moved the shift from 0 to 10 to recieve a sharper image.</p>
 106
 107 <img src="jotunheimen-std-ir-piece.jpg" align="left" width="25%">
 108 <img src="jotunheimen-std-ir-pieceimg.jpg" align="right" width="40%">
 109
 110 <br clear="all">Next we tried the piecewise linear stretching for
 111 contrast. In this image we tried to make all of the histograms in the
 112 red, blue and green spectrum as similar as possible so we could detect
 113 a change in the image.(insert histogram change)
 114
 115 <img src="jotunheimen-std-ir-piece2.jpg" align="left" width="25%">
 116 <img src="jotunheimen-std-ir-pieceimg2.jpg" align="right" width="40%">
 117
 118 <br clear="all">We tried to break the slope and move the break point
 119 to slightly after each histogram peak. This resulted in the image
 120 obtaining a slightly blue tint and dullness. (put ugly blueish picture
 121 here) As this result was not really increasing the contrast, we tried
 122 another variation to try to spread out the histogram peak to use a
 123 wider range.  This setting gave an improved image, were it is easier
 124 to see the red vegetation and the white ice.</p>
 125
 126 <br clear="all"><img src="jotunheimen-std-ir-eq.jpeg" align="right" width="40%">
 127 <p clear="all">We also tried to do histogram equilization on the
 128 standard infrared composition.  This changed the colours in the image,
 129 making the previously green areas red, and the brown areas more light
 130 blue.  In this new image, we can clearly see the difference between
 131 two kind of water, one black and one green.  We suspect the green
 132 water might be deeper, but do not know for sure.</p>
 133
 134 <p>We can get best contrast stretch by using the histogram
 135 equalisation.  This gave us the widest range of visible separation
 136 between features.
 137
 138 <br clear="all">
 139 <h3>Displaying colour images</h3>
 140
 141 <p><img src="jotun2000topper.jpg" width="40%">
 142 <!-- img src="jotunheimen-map.jpeg" -->
 143
 144 <img src="jotunheimen-std-ir.jpeg" width="40%"></p>
 145
 146 <p>Comparing a map we found on the web, and the standard infrared
 147 image composition, we can identify some features from the colors
 148 used:</p>
 149
 150 <img src="jotunheimen-ir-2band.jpeg" align="right" width="40%">
 151 <ul>
 152
 153  <li>water is black or green
 154
 155  </li><li>ice and glaciers are white, while snow is light green.
 156
 157  </li><li>vegetation is red.
 158
 159  </li><li>non-vegetation is brown or dull red when closer to snow and
 160    glaciers.
 161
 162 </li></ul>
 163
 164 <p>Next, we tried to shift the frequencies displayed to use blue for the
 165 red band, green for the near ir band and red for the mid ir (1.55-1.75
 166 um).  With this composition, we get some changes in the colours of
 167 different features:</p>
 168
 169
 170 <ul>
 171  <li>water is black
 172
 173  </li><li>ice and glaciers are light blue, while snow is dark blue.
 174
 175  </li><li>vegetation is light green and yellow.
 176
 177  </li><li>non-vegetation is red or brown.
 178
 179 </li></ul>
 180
 181
 182 <h2>Filtering and image sharpening</h2>
 183
 184 <img src="jotunheimen-band4.jpeg" width="40%">'
 185 <p clear="all">We decided to work on the grey scale version of the
 186 near infrared (band4).  We changed the colour assignment to use this
 187 band for all three colours, giving us a gray scale image.</p>
 188
 189 <img src="jotunheimen-band4-low3.jpeg" width="40%">'
 190 <p clear="all">We applied the 3x3 low pass filter on this image, and
 191 this gave us almost the same image as the original.  If you look
 192 closely you can see that some white dots in the original disapper, and
 193 some of the water edges seem to blur very slightly.</p>
 194
 195 <img src="jotunheimen-band4-high3.jpeg" width="40%">'
 196 <p clear="all">We also tried the 3x3 high pass filter on the band4
 197 grey scale image.  This gave a very noisy image.  Edges of vallies and
 198 ice are not well defined.  The black waters are still obvious.</p>
 199
 200 <img src="jotunheimen-band4-edge3.jpeg" width="40%">'
 201 <p clear="all">We also tried the 3x3 edge detection, and this gave us
 202 an image that makes it difficult to distinguish elevation features
 203 such as the valleys. Rather, edge detection allows us to study main
 204 features in an area like the lakes. (insert band4 edge 3 image)
 205
 206 <img src="jotunheimen-band4-grad3.jpeg" width="40%">'
 207 <p clear="all">We tried a gradient filter using this 3x3 matrix.  The
 208 matrix was chosen to make sure the sum of all the weights were zero,
 209 and to make sure the sum of horizontal, vertical and diagonal numbers
 210 were zero too.</p>
 211
 212 <p><table align="center">
 213   <tbody><tr><td>1</td><td>2</td><td>-1</td></tr>
 214   <tr><td>2</td><td>0</td><td>-2</td></tr>
 215   <tr><td>1</td><td>-2</td><td>-1</td></tr>
 216 </tbody></table></p>
 217
 218 <p>The gradient filter used gave us enhancement on lines in the
 219 vertical, horizontal and diagonal directions. This is seen by the
 220 white lines that outline certain areas of main features like the
 221 rivers within the vallies and some of the lakes.</p>
 222
 223 <img src="jotunheimen-band4-neg1.jpeg" width="40%">'
 224 <p>When we rework the matrix to equal negative one, we end up with a
 225 lot of noise in the image that also seems to blurr the image. Using a
 226 negative one matrix is not optimal if you are trying to obtain
 227 sharpness.</p>
 228
 229 <p><table align="center">
 230   <tbody><tr><td>-1</td><td>-1</td><td>-1</td></tr>
 231   <tr><td>-1</td><td>7</td><td>-1</td></tr>
 232   <tr><td>-1</td><td>-1</td><td>-1</td></tr>
 233 </tbody></table></p>
 234
 235 <img src="jotunheimen-band4-plus1.jpeg" width="40%">'
 236 <p clear="all">We then tried with a 3x3 matrix were the sum of all
 237 values equals 1, to enhance the high frequency parts of the image.</p>
 238
 239 <p><table align="center">
 240   <tbody><tr><td>-1</td><td>-1</td><td>-1</td></tr>
 241   <tr><td>-1</td><td>9</td><td>-1</td></tr>
 242   <tr><td>-1</td><td>-1</td><td>-1</td></tr>
 243 </tbody></table></p>
 244
 245 <p clear="all">This gave us a sharper looking image compared to the
 246 result of the negative 1 filter.  This is not really obvious unless
 247 one is comparing the two images carefully.  In order to see more
 248 differences the matrix sums would have to be more then plus/minus one.</p>
 249
 250 <h2>References</h2>
 251
 252 <ul>
 253   <li><a href="http://www.cs.uu.nl/wais/html/na-dir/sci/Satellite-Imagery-FAQ/part3.html">Satellite-Imagery-FAQ</a>
 254 </li></ul>
 255
 256     <hr>
 257     <address><a href="mailto:pere@hungry.com">Petter Reinholdtsen</a></address>
 258 <!-- Created: Sun May  1 13:25:38 CEST 2005 -->
 259 <!-- hhmts start -->
 260 Last modified: Sun May  1 14:28:48 CEST 2005
 261 <!-- hhmts end -->
 262   </body>
 263 </html>