4 Progressive Techniques
67
raster data and include interleaving techniques and other mechanisms based on image compression. The success and application of image compression mechanisms for
progressive raster transmission is due to their effectiveness: they provide good compression with low loss of information. Furthermore, they are relatively easy to implement. The main characteristics of these approaches are summarized in Sect. 4.2.1.
In the vector domain the most successful methods for progressive transmission
are restricted to the distribution of data in the form of triangular meshes. Some of the
proposed approaches are discussed in Sect. 4.2.2.
4.2.1 Progressive Transmission of Raster Images
The first successful implementations of progressive raster transmission rely on interleaving techniques. The simplest approach consists of randomly extracting subsets of pixels from the image and incrementally completing it by adding pixels.
If the implicit row/column ordering of images is exploited, pixels can be sampled
in a uniform fashion; alternatively, hierarchical structures, such as quadtrees [45],
can be used to extract more pixels from parts of the image with higher density
of detail.
Sophisticated compression techniques have generated more effective progressive
raster transmission methods (see, for example, [41, 47]). The most commonly used
image compression method is the JPEG (joint photographic experts group) format.
The JPEG method is a transform-based compression method that decomposes the
original data before compression (see [23] for a detailed description). As images in
JPEG format are segmented into rectangular sub-blocks and each block is transformed independently, at high compression ratios they take on unnatural blocky
artifacts.
Alternative techniques are based on wavelet decompositions [33]. Wavelet methods are also transform-based but they act on the entire image. The result is a robust digital representation of a picture, which maintains its natural look even at high
compression ratios (see [12] for an overview). Wavelets are well suited to progressive transmission because, besides providing more efficient overall compression, they
naturally represent image data as a hierarchy of resolution features and its inverse at
each level provides subsampled versions of the original image. Therefore, progressive transmission corresponds to a natural reconstructive mode for a wavelet-based
compression algorithm (see, for example, [41]).
Much work on image compression relies on fractal theory [3]. A fractal is a
geometrical figure whose local features resemble its global characteristics (selfsimilarity property). The challenges in fractal-based compression include finding a
small number of affine transformations to generate the image as well as subparts
of the input image that have self-similarity properties [13, 21]. Hybrid compression
methods, combining fractals and wavelets, have also been defined [14, 50].
When transform-based methods are used to compress an image, during transmission over the Internet, instead of the image, the function coefficients are transmitted,
and the image is subsequently reconstructed by inverting the transformation. Both
67
raster data and include interleaving techniques and other mechanisms based on image compression. The success and application of image compression mechanisms for
progressive raster transmission is due to their effectiveness: they provide good compression with low loss of information. Furthermore, they are relatively easy to implement. The main characteristics of these approaches are summarized in Sect. 4.2.1.
In the vector domain the most successful methods for progressive transmission
are restricted to the distribution of data in the form of triangular meshes. Some of the
proposed approaches are discussed in Sect. 4.2.2.
4.2.1 Progressive Transmission of Raster Images
The first successful implementations of progressive raster transmission rely on interleaving techniques. The simplest approach consists of randomly extracting subsets of pixels from the image and incrementally completing it by adding pixels.
If the implicit row/column ordering of images is exploited, pixels can be sampled
in a uniform fashion; alternatively, hierarchical structures, such as quadtrees [45],
can be used to extract more pixels from parts of the image with higher density
of detail.
Sophisticated compression techniques have generated more effective progressive
raster transmission methods (see, for example, [41, 47]). The most commonly used
image compression method is the JPEG (joint photographic experts group) format.
The JPEG method is a transform-based compression method that decomposes the
original data before compression (see [23] for a detailed description). As images in
JPEG format are segmented into rectangular sub-blocks and each block is transformed independently, at high compression ratios they take on unnatural blocky
artifacts.
Alternative techniques are based on wavelet decompositions [33]. Wavelet methods are also transform-based but they act on the entire image. The result is a robust digital representation of a picture, which maintains its natural look even at high
compression ratios (see [12] for an overview). Wavelets are well suited to progressive transmission because, besides providing more efficient overall compression, they
naturally represent image data as a hierarchy of resolution features and its inverse at
each level provides subsampled versions of the original image. Therefore, progressive transmission corresponds to a natural reconstructive mode for a wavelet-based
compression algorithm (see, for example, [41]).
Much work on image compression relies on fractal theory [3]. A fractal is a
geometrical figure whose local features resemble its global characteristics (selfsimilarity property). The challenges in fractal-based compression include finding a
small number of affine transformations to generate the image as well as subparts
of the input image that have self-similarity properties [13, 21]. Hybrid compression
methods, combining fractals and wavelets, have also been defined [14, 50].
When transform-based methods are used to compress an image, during transmission over the Internet, instead of the image, the function coefficients are transmitted,
and the image is subsequently reconstructed by inverting the transformation. Both