Data Transmission and Decompression

Figure 5.4: Estimated gradients (a) are used for shading until the original gradient data has arrived (b)

(a)	(b)

The compressed data set consists of two parts: a header, which contains control-information about the objects and their position within the data, information about additional data channels and how to use them for rendering. The body contains voxel positions and other data channels for all objects. The data within the body is arranged in a way which allows to obtain a view on the data as early as possible during loading. Objects and data channels which are more significant for the preset visualization mappings are stored and transferred earlier than less significant data (like actual gradient information). Data channels are subdivided into blocks of a few Kilobytes each. As soon as an entire block has arrived, it can be decompressed and displayed while the following data is arriving. This allows voxel data to be rapidly updated, without having to wait for the arrival of the entire channel. Finally, as gradient information usually accounts for most of the data to be transmitted (See table 5.1), for boundary objects a locally computed gradient approximation (linear regression [45] with a filter size of 5 while interpreting the data as a binary object) can be displayed before the original gradient data arrives (see figure 5.4). For inherently binary objects, like basins of attraction within the phase space of a dynamical system [4] the locally computed gradients can entirely replace the transmission of gradients, significantly decreasing the amount of transmitted data.

mailto:mroz@cg.tuwien.ac.at.