Page 132 - Digital Analysis of Remotely Sensed Imagery

Storage of Remotely Sensed Data 103

are allowed, up to and including the entire image (i.e., no tiles). Each
tile must be of the same size except border tiles. Treated as a separate
image, each tile is processed independently using the one-dimensional
discrete wavelet transform. It can be decomposed to different levels,
to which different coefficients in quantization are applied. It can also
be accessed and referenced independently of each other. Thus, it is
possible to uncompress any portion of the compressed image.
The JPEG 2000 image compression process is made up of several
steps that may include data ordering, arithmetic entropy encoding,
coefficient bit modeling, quantization, wavelet transformation of
tiles, level shifting and component transformations, and coding of
images with regions of interest. Once the entire image has been
compressed, a postprocessing operation passes over all the compressed
blocks and determines the extent to which each block’s embedded bit
stream should be truncated to achieve a desired bit rate, distortion
bound, or other quality metric. A separate bit stream is generated for
each tile. The bit stream is organized as a succession of layers, where
each layer contains the additional contributions from each code tile
(some contributions may be empty). The final bit stream is composed
of a collection of such layers. Each layer has an interpretation in terms
of overall image quality, indicating the discrete lengths to which the
bit stream has been truncated.
To decode the compressed image, the compression procedure is
reversed. The inverse discrete wavelet transformation is applied to
the compressed image file, undo the coefficient bit modeling, undo
the entropy encoding, and read the tiles from the codestream (Miljour).
The information contained in the tiles and marker headers tells the
decoder how to reconstruct the original image. Since images no
longer need to be divided into subimages of 8 8 pixels, the artifacts
of uncompressed JPEG images are avoided.
In additional to the above compression methods, there is another
method called fractal compression. This method demands a huge
amount of time to generate fractal formulae from an image. However,
the reverse process is very simple and can be achieved relatively fast.
The compression ratio that can be accomplished with fractal
compression is independent of zooming and asymmetric process.
This technique of compression is very suitable for applications where
compression is done at a single location and decompression at a large
number of places, such as distribution of remotely sensed data.
Compared with the above techniques, the quality of a fractal-
decompressed image is independent of zooming, which makes it
suitable for graphics applications. In fractal transform, the entire
image is presented in terms of parts of itself and encoded. Fractal
basics can be described as a notion of “futuristic photocopies.”
The amount of information lost through the compression-
uncompression process is dependent upon the time spent in deriving
the fractal formulae. The more time spent in the encoding process, the