start
contents
Abstract
Most image processing techniques involve looking at a pixel and calculating a new intensity value based upon its neighbouring pixel values. If a pixel is chosen at random, most of the time it is expected that the neighbours of that pixel will be part of the same region. However, sometimes the chosen pixel will be on the edge of a region and so neighbouring pixels may have values from a differing region. It's these values that should be excluded from calculations.
In recent times, more effort has been made to investigate and produce image processing techniques that help deal with this fundamental problem. This thesis focuses on providing a method of deciding whether a group of pixels is in the middle of a region or on the edge of a region. These methods are then applied to image filtering algorithms where it will be shown that local segmentation can greatly reduce the amount of structure lost during filtering, and compares favorably to state of the art image filters.
Acknowledgements
I would like to acknowledge my project supervisor Dr Peter Tischer, for letting me investigate his idea of local segmentation.
I would like to thank Dr Henry Wu for providing me with images used by Tao Chen in his paper testing the Tri-State Median Filter.
start
contents