Image Overviews, Tiling and Pyramids - Open Source Geospatial Tools: Applications in Earth Observation - page 84

Chapter 7
Image Overviews, Tiling and Pyramids

Image overviews, tiling and pyramids are techniques to view large images more
efficiently. In this chapter we explain each of them, based on the utilities
gdaltindex , gdaladdo , gdal_retile.py and gdal2tiles.py .
An image overview is a downsampled version of the original image. The overviews
can be located in external files or, for some image formats, be included within the
image file itself. Image tiles are subsets of a larger image at the same level of detail.
We have already seen in Chap. 3 how tiling can be used to encode raster data in a file.
In this context, tiling can also be a method to reduce the file size of an image that
covers a large geographical area. Each tile is then stored in a separate file that can be
accessed more efficiently for a certain region of interest. Building image pyramids
of a raster image is a technique where image overviews are combined with tiling.
The pyramid in Fig. 7.1 represents a raster dataset with 4,096

4,096 pixels at
four different scales. At each scale, the image is divided in a number of tiles. Each
tile has the same number of pixels (256

×

256 in this example). At the original spatial
resolution (level 0), a maximum of tiles is needed to represent the entire image. At
the coarsest resolution (level 4), the image can be represented by a single tile. The
tiles are numbered following the scheme in Sect. 7.3 .
Especially if a large archive of images needs to be accessed through the Inter-
net, building multi-scale image representations can considerably increase the perfor-
mance. Subsets of images can be accessed more efficiently by selecting individual
tiles rather than a single file of a larger size. When raster data covering a large area
are displayed on screen, it is not needed to show all available detail at once. From
a distance there are not sufficient pixels to display the data at full resolution. Only
as a user zooms in some region of interest, the tiles are presented in increasing level
of detail. This technique is also used by Google for its virtual globe Google Earth.
GDAL includes a number of utilities to support image overviews, tiling and building
pyramids, which are described in this section.

×