My Phd work:

I defended my PhD on September 25th, 1999, and received my PhD on 25th November 1999, from University of Oulu, Finland. The PhD thesis was entitled "Accelerating Direct Volume Rendering using Seed Filling in the View Lattice", and the following presents a short abstract. For more information, take a look at my publications.


Three-dimensional visualization of volumetric images has emerged as a widely-used method for visualizing tomographic MR or CT images. It has been used in image-guided surgery (IGS) systems, including neurosurgery, orthopaedics, ENT (ear, nose and throat) and spine procedures. The interactive visualization of these data sets is becoming possible due to the fast development of computer hardware and volumetric visualization algorithms.

This thesis concentrates on fast volume visualization algorithms, and in particular speeding up the rendering process by reducing the number of accesses to the sample points in the view lattice.

There are three considerations that each direct volume rendering algorithm should address. The first is the format of the view lattice, i.e. how are the sample points located. The second is the order in which the sample points are referenced and the third is the shading model applied to get the final image.

Two different view lattices have been used in this work. The lattices are referred to as the Bresenham-style lattice and the template-style lattice. They are named after the primitives that are used to move from one sample point to the neighboring sample point.

This thesis proposes the use of a seed fill algorithm to decide the order in which the sample points in the view lattice are referenced. In particular we have presented three seed fill algorithms that all reduce the number of references to the sample points that do not belong to the object that is being displayed. This is the main contribution of this work.

The rendering quality is also addressed by combining methods for higher image resolution and interactive light model adjustment with the seed filling algorithms. Image resolution is further improved by using shape-based interpolation and multiple templates. The light model may be modified interactively, because we encode the normal direction into the volumetric data set. Thereby all visible voxels can be run through a lookup table that is modified for each frame with the current light settings.

The seed filling algorithms presented are compared with other volume rendering techniques, such as ray casting, proximity cloud optimization and the plain template technique. It is shown that the algorithms may be combined with both the proximity cloud optimization and template techniques, hence improving the performance of all algorithms.

This research has developed and evaluated new acceleration techniques for volume rendering. The techniques presented are applicable in many volume rendering applications, from which our main interest has been image-guided surgery and MRI-guided interventional radiology.