Chapter 7: Raster Data

1
Chapter 7Raster Data
What is this?

• Paul Sutton
• psutton_at_du.edu
• Department of Geography
• University of Denver

2
The Raster Data Model

• The raster data model is more like a photograph
• than a map. If you look at a photograph through
• a strong magnifying glass, youll see that it is
• made up of a series of dots of different colors
  or
• shades of gray. The raster data model works in a
• similar way it is a regular grid of dots
  (called
• cells, or pixels) filled with values. In fact,
  when
• a picture is stored in a computer, the raster
  data
• model is used.
• In the photo above, you can see roads, cars,
  boats, and a river. Notice that there are
• no boundaries drawn on the photograph to
  distinguish features it is a continuous
• surface. Using the raster data model, the earth
  is treated as one continuous surface.

3
Raster or Grid Structure

• In the raster data model, each location is
  represented as a cell. The matrix of cells,
  organized into rows and columns, is called a
  grid. Each row contains a group of cells with
  values representing a geographic phenomenon. Cell
  values are numbers, which represent nominal data
  such as land use classes, measures of light
  intensity or relative measures.

4
Interpreting a Raster

• There are three ways to interpret each dot in a
  photograph. The first is to classify each dot as
  belonging to something - a group of similarly
  classified pixels becomes an object, Like
  Forest. The second way of interpreting is
  simply to measure the value of its color or shade
  of gray some pixels would be black, some pixels
  would be medium gray, etc. The third way is to
  define the pixel relative to a known reference
  point, such as mean sea level (for elevation) or
  distance from Graceland.
• The same three interpretations can be used for
  the raster data model in GIS. The cell value can
  represent a classification, such as vegetation
  type. It can be a measurement, such as a
  satellite measuring the amount of light reflected
  by the earth. Finally, it can be an
  interpretation of elevation.

5
How do these Images of Lincoln Differ?
6
Points, Lines, Polygons as Grids
7
How the Raster Stores Spatial Relationships

• Because the raster data model is a regular
• grid, spatial relationships are implicit.
• Therefore, explicitly storing spatial
• relationships is not required as it is for the
• vector data model.
• Notice that in a grid, cells have eight (8
  neighbors (except those on the
• outside edges) four off the corners and four off
  the sides. Cells are
• identified by their position in the grid. In the
  example above, with an origin
• in the upper-left, the cell (3,2) would be 3 over
  along the x-axis and 2 down
• on the y-axis. Finding any one of eight neighbors
  simply requires adding or
• subtracting one (1) from the x or y values. For
  example, the value to the left
• of (3,2) is (3-1,2) or (2,2).

8
Geo-Referencing in Raster

• Raster data is geo-referenced by
• specifying the coordinate system to
• which a grid is registered, the real-
• world location of the reference
• point and the cell size in real-world
• distances. Typically, the upper-left
• or the lower-left corner of the grid
• is used as the reference point. This
• reference point location, along with
• the cell size, can be used to
• determine the geographic location
• of any cell within the raster data
• set. Using the same coordinate
• system, raster data sets can be
• logically organized into subjects for
• geographic analysis.

9
How the Raster Data Model Represents Surfaces

• In representing surfaces, the surface
• value (e.g., elevation) is recorded
• for each cell. Rather than having
• the value represent the entire cell,
• it represents only the center point
• of the cell. This set of center points
• for cells in a grid is called a lattice.
• The lattice supports accurate surface
• calculations. The types of surface
• calculations used for analysis include
• slope (the rate of change in the value
• over distance), aspect (direction the
• slope faces), and contour interpolations.
• The United States Geological Survey
• (USGS) uses this model to create their
• Digital Elevation Model (DEM) product.

Measures are taken at lattice point (each
cross). Isolines (in this case, contours) can
be interpolated from the lattice.
10
Types of Raster Data

• Remotely Sensed Imagery
• Landsat, MSS
• POES AVHRR DMSP
• SPOT
• Digital Elevation Models (DEMs)
• Digital Orthophotos (DOQs)
• Binary Scanned Files
• Digital Raster Graphics (DRGs) (Scanned topo
  maps)
• Formats TIFF, GeoTIFF, BIL, BSQ, GIF, JPEG, etc

11
Landsat Thematic Mapper

• This is a Landsat TM image of the Grand Canyon.
  Landsat has seven bands, 30 meter resolution and
  global coverage. It is a major instrument used in
  global change studies including studies of fires
  and deforestation.

12
SeaWIFs Dust Storm over Koreas and Northeast US

• This pair of true-color images
• was collected by the Sea-viewing Wide
  Field-of-view
• Sensor (SeaWiFS). The top image, acquired
  April 22, 2001,
• shows haze over the mid-Atlantic United
  States extending
• well out over the Gulf Stream. The bottom
  image, acquired
• April 24, reveals that a significant
  amount of dust is still
• blowing out of Asia.

13
3-D Visualizations Using DEMs
14
AVHRR Image of Los Alamos Fire

• AVHRR (Advanced very High Resolution Radiometer
  is a weather satellite run by NOAA with 1 km
  spatial resolution

15
DMSP image of the Earth at Night

• How was this image made?

16
Digital OrthoPhotoQuad (DOQ) of Washington DC area
17
IKONOS 1 meter resolution Image of the U.S.
Mexico border in Nogales AZ/Sonora
18
Cool 3-D Applications for DEMs
19
Types of overlay operations that can be performed
on Raster Data

• And
• Or
• Max
• Min
• Exhaustive set

20
Overlay Can you imagine overlay logic for the
operation below? (e.g. A combined with B to
produce C)
A B C
21
Buffer
1
22
Complex Retrieval Map Algebra

• Combinations of spatial and attribute queries can
  build some complex and powerful GIS operations,
  such as weighting.

23
Recode
OR
24
Describing a classed raster grid
20
P (blue) 19/48
15
10
5
25
The mixed pixel problem
26
Rasters and vectors can be flat files if they
are simple
Flat File
Vector-based line
4753456 623412
4753436 623424
4753462 623478
4753432 623482
4753405 623429
4753401 623508
4753462 623555
4753398 623634
Raster-based line
Flat File
0000000000000000
0001100000100000
1010100001010000
1100100001010000
0000100010001000
0000100010000100
0001000100000010
0010000100000001
0111001000000001
0000111000000000
0000000000000000
27
Data Compression Tricks

• Run Length Encoding
• Quad-Tree Structure

28
Run Length Encoding

• The run-length encoding (RLE) scheme can be
  used on any kind of data, but works best when the
  data contains repeating strings, or runs, such as
  the same character or consecutive pixels of the
  same color. In its simplest form, RLE involves
  scanning a line of text and converting strings of
  consecutive like characters into two-byte
  packets, the first bite containing the integer
  number of repetitions, or run count and the
  second bite containing the character, or run
  value. For example, the 31-byte string
• RRRRRSSSSSSSSSSTTTTTTUUVVVWXXXX
• would be represented as the 14-bite string
• 5R10S6T2U3V1W4X ,
• a compression ratio of 21. Each time the
  character changes, a new packet is generated.
  Although this method implements easily and
  executes rapidly, it is not always efficient.

29
The quad-tree structureA spatial index which
recursively decomposes a dataset (grid) into
square cells of different sizes until each cell
has a homogenous value
0
1
2
3
210
1
0
quadrant
3
2
number
Figure 3.9
The quad-tree structure. Reference to code 210.
30
Vector to raster exchange errors
31
Projection Geometric Transformation of Raster
Data

• Ground Control Points
• Warping or Rubber-Sheeting
• Resampling
• Nearest Neighbor
• Bilinear Interpolation
• Cubic Convolution

32
Resampling
33
Nearest Neighbor Resampling

• The nearest neighbor approach uses the value of
  the closest input pixel for the ouput pixel
  value. To determine the nearest neighbor, the
  algorithm uses the inverse of the transformation
  matrix to calculate the image file coordinates of
  the desired geographic coordinate. The pixel
  value occupying the closest image file coordinate
  to the estimated coordinate will be used for the
  output pixel value in the georeferenced image.
• ADVANTAGES
• Output values are the original input
  values. Other methods of resampling tend to
  average surrounding values. This may be an
  important consideration when discriminating
  between vegetation types or locating boundaries.
  Since original data are retained, this method is
  recommended before classification.
• Easy to compute and therefore fastest to
  use.
• DISADVANTAGES
• Produces a choppy, "stair-stepped" effect.
  The image has a rough appearance relative to the
  original unrectified data. Data values may be
  lost, while other values may be duplicated.
  Figure 1 shows an input file (orange) with a
  yellow output file superimposed. Input values
  closest to the center of each output cell are
  sent to the output file to the right. Notice that
  values 13 and 22 are lost while values 14 and 24
  are duplicated.

34
Graphic Explanation of Nearest Neighbor
Resampling
35
Bilinear Interpolation

• The bilinear interpolation approach uses the
  weighted average of the nearest four pixels to
  the output pixel.
• ADVANTAGES
• Stair-step effect caused by the nearest
  neighbor approach is reduced. Image looks smooth.
• DISADVANTAGES
• Alters original data and reduces contrast
  by averaging neighboring values together.
• Is computationally more expensive than
  nearest neighbor.

36
Cubic Convolution

• The cubic convolution approach uses the weighted
  average of the nearest sixteen pixels to the
  output pixel. The output is similar to bilinear
  interpolation, but the smoothing effect caused by
  the averaging of surrounding input pixel values
  is more dramatic.
• ADVANTAGES
• Stair-step effect caused by the nearest
  neighbor approach is reduced. Image looks smooth.
• DISADVANTAGES
• Alters original data and reduces contrast
  by averaging neighboring values together.
• Is computationally more expensive than
  nearest neighbor or bilinear interpolation.