jnovatnack.com

I successfully defended my master's thesis titled Scale-Dependent/Invariant Local 3D Geometric Features and Shape Descriptors. It was a day I have been looking forward to for awhile and I am proud and happy with the final product of my research. Thanks to everyone for coming!

I am happy to say a paper titled A Generalized Family of Fixed-Radius Distribution-Based Distance Measures for Content-Based fMRI Image Retrieval will be published in an upcoming edition of Pattern Recognition Letters. I am a co-author of the paper along with a number of colleagues at Drexel University, the University of Toronto and Rutgers University (Nicu Cornea, Ali Shokoufandeh,

I also have a personal (non-technical) blog.

PageRank for Product Image Search describes a method of applying the PageRank algorithm to sets of images to improve image search results from text based image search engines, i.e. Google images. Unlike the PageRank algorithm traditionally used by Google, where edges represent hyperlinks manually placed between images or web pages, the graph represents visual relationships between the images. These visual relationships are determined by comparing appearance models representing each image.