Volume visualization is the technique of displaying two-dimensional projections of three-dimensional (volume) data.
Volume rendering techniques of 3D data are classified as direct and indirect. Indirect volume rendering methods,
such as Marching Cubes , extract intermediate geometric meshes of iso-surfaces from volume data and render
them using surface rendering methods. Indirect methods are more suitable for applications where the visualization of
surfaces is important. Direct Volume Rendering (DVR) techniques render the volume data without generating any
intermediate representation; thus facilitating the visualization of the material inside. By specifying a proper transfer
function in the DVR process, voxels are assigned with certain optical properties and different structures are revealed
in the resulted images.
The techniques of DVR are useful for various kinds of scientific visualization, such as medical diagnosis and life
science research. However, structures may not be clearly shown in the resulted images, due to various reasons like
poor lighting and reflection parameters. Thus, the pixel values cannot provide enough implication on existences of
structures. For example, a homogeneous region in a DVR image may represent the fine details of a structure. The
image should have displayed a variation to indicate this feature. Meanwhile, many data processing operations may
have been applied to the data set before the volume rendering process taken place, including filtering, denoising,
simplification, watermarking, or compression. These operations introduce slight modifications on the shape of
interesting features hidden inside the data volume and modify their visual appearances. Even if the pixel values do
reveal variations, the viewer may fail to detect the existences of structures. This is because that a human observer
excels at pattern recognition and classfication tasks but often poor at detailed scoring tasks. It is common for the
human observer to miss objects, score objects twice, and produce unsteady traces . Indeed, it is not un-usual for
the same observer to produce different markups for the same image at...