Page 122 - Digital Analysis of Remotely Sensed Imagery

Storage of Remotely Sensed Data 93

placement and extents of the image to follow, and the pixel display
sequence. As with all raster data, pixels in a row are stored left to
right sequentially. The entire image is stored row by row sequentially
from top to bottom.
GIF uses a palette of only 256 colors that allows a single band of
continuously varying tone (i.e., grayscale) to be represented
adequately if it is recorded at 8 bits such, as with Landsat Thematic
Mapper (TM) imagery. However, the tens of thousands of colors on a
color composite of three spectral bands would suffer considerably in
its quality if stored in the GIF format. They require 24 bits to store
faithfully, 8 bits for each band, creating a rather large 256-color file.
The GIF format was thus improved in 1989 to include a compression
function. The lossless LZW compression (see Sec. 3.4.3) was adopted
to reduce image size without degrading image quality. GIF files offer
optimum compression (e.g., smallest files) for solid-color graphics.
Initially designed for the explicit purpose of uploading graphics
to the World Wide Web, GIF is particularly suited to store images that
are captured using screen dumping and that are to be embedded into
other systems such as Word and PowerPoint for crude visualization.
In this sense, it is useful for capturing digitally processed remote
sensing results for presentation at professional meetings. This format
is not suited to store either raw or processed remote sensing data, or
any graphic results derived from them.
3.3.3 JPEG
Named after the group that originated it, the JPEG format is a popular
and efficient graphic format for storing images, albeit not always
faithfully. Frame images of continuous tone in binary, grayscale, or
color can be stored in JPEG. This format is particularly suited for those
images that must be reduced to a very small size through image
compression. There are three coding systems (Gonzalez and Woods,
2002):

• A lossy baseline coding system that is adequate for most
compression needs
• An extended coding system for greater compression
• A lossless independent coding system for reversible compression

During data compression, image pixel values are modified via a
complex mathematical formula. The relationship between a pixel and
its neighbors is examined in all directions to identify the factors for
that formula so that these pixels can be best represented. Minor detail
that does not fit for compression is not retained in the compressed
image in order to achieve a high efficiency of data compression at the
expense of losing image quality.
Prior to storing an image in the JPEG format, the system analyst
is given the option of specifying the amount of compression desired,