VOLUMETRIC IMAGE ALIGNMENT UTILIZING PARTICLE SWARM OPTIMIZATION

摘要

We have presented a method for aligning stacks of images and stitching them to form a single, large-scale volumetric data set. The method is able to produce not only an alignment score but also a confidence score, which is an indicator of how likely the image stack tiles are correctly aligned. This allows the user to quickly identify where the alignment had low confidence levels and manually align only those portions of the image. Further research needs to be done to study how ambiguous solutions can be addressed. This may be done by cross-correlating information from multiple planes or by using alternative registration techniques.