Remote Sensing


Geog 483/553
Fall 2011 
Tu Th 12:30am - 1:50pm
352 Fillmore 
Instructor: Ling Bian 
Office: 120 Wilkeson Quad
Office hours: Tu Th 2-3pm or by appt
TA:  Steve Tulowiecki
Lab Tu  6:30-7:50pm, W145
       Thur 5:00-6:20pm,  W145   

Image  Enhancement
    It increases distinction between features in a scene

1. Contrast manipulation
   Gray-level threshold
        - segmenting an image into two classes - binary mask
        - additional processing can be applied to each segment
 

   Level slicing
        - dividing the histogram of DNs into several slices
        - extensively used in thermal images to show temp ranges
 

   Contrast stretching
        - expanding a narrow range of DNs to a full range
                                              DN - Min
        Linear stretch: DN' = (--------------) 255
                                             Max - Min
 

        - advantage: simple computation and efficient storage
        - disadvantage: rare and frequent values have the same amount of levels

    Histogram-equalized stretch
        - stretching based on frequency of occurrence
        - frequently occurred DNs have more display levels
 

        - special stretch
 
 

2. Spatial feature manipulation
   Spatial filtering
        - Low pass filters emphasize low frequency features
        - computing the average values of moving windows
 

        - High pass filter emphasizes local details
        - subtracting a low-pass filter from the original image
 

   Convolution
        - a moving kernel with a weighting factor for each pixel
 

   Edge enhancement
        - adding back the original image to the high frequency image component
        - preserving both the original and the high frequency features

   Directional first differencing
        - displaying the differences in gray levels of adjacent pixels

        - the direction can be horizontal, vertical, or diagonal

        - it is necessary to add a constant to the difference for display purposes

           Contrast stretching is needed for all feature manipulations
 
 

3. Multi-image manipulation
   Spectral ratioing
        - a ratio of two bands (with great difference in reflectance)
 

        - useful to eliminate effects of illumination differences
 

        - necessary to stretch the resultant values to a full range of DN after ratioing
 

   Hybrid color ratio composite
        - problem: different features of similar ratio may appear identical

        - solution: combining 2 ratio bands + 1 original band to restore the absolute DN
 
 

    Principle component transformation
        - to reduce redundancy in multi-spectral data

        The transformation:
                                        DNI  = a11DNA + a12DNB
                                        DNII = a21DNA + a22DNB

        DNI, DNII = DNs in new component images
        DNA, DNB = DNs in the original images
        a11, a12, a21, a22 = coefficients for the transformation

        - after the axes rotation, the original n bands images are
            converted into n principle components images

        - the first component (PC1) image contains the largest
            percentage of the total scene variance (90%±)

        - the second component (PC2) contains the largest % of the
            remaining variance

        - each succeeding component contains decreasing amount of
            variance

        - successive components are orthogonal; they are not
            correlated to each other

        - PCs can be used as new bands for image classification

        - PCA is scene specific
 

    Kauth-Thomas tasseled cap
        - an orthogonal transformation

        - the 4 MSS bands can be converted into 4 new bands:
          brightness, greenness, yellow stuff, and non-such

        - the first two indices contain the most info (90+%)

        - brightness is related to bare soils

        - greenness is related to the amount of green vegetation

        - the 6 TM bands can be converted into a 3-D space:
          plane of soils, plane of vegetation, and a transition zone

        - a third feature, wetness

        - the K-T transformation is transferable between scenes
 
 

    Intensity-Hue-Saturation transformation (IHS)
        transform the RGB space into the IHS space for color enhancement

        - intensity: brightness, hue: color, saturation: purity

        - the hexcone model projects the RGB cube to a plane, resulting in a hexagon

        - the plane is perpendicular to the gray line and tangent to
            the cube at the "white" corner

        - intensity = distance along the gray line from the black point to any given hexagonal projection

        - hue = angle around the hexagon

        - saturation = distance from the gray point at the center of  the hexagon

        - I,H,S = f(R,G,B), I' = f(I+Ipan), H' = f(H+Hpan), S' = f(S+Span)
          R',G',B' = f(I',H',S')
 
 
 

        - IHS is flexible in image enhancement
 

4. Reading chpt 7
 
 
 
 

...Back to Ling Bian top page.