PROJECT TITLE :

Optimized Update/Prediction Assignment for Lifting Transforms on Graphs - 2018

ABSTRACT:

Transformations on graphs will give compact representations of signals with many applications in denoising, feature extraction, or compression. In explicit, lifting transforms have the advantage of being critically sampled and invertible by construction, however the potency of the transform depends on the choice of a smart bipartition of the graph into update (U) and prediction (P) nodes. This is the update/prediction (U/P) assignment drawback, that is the main target of this Project. We have a tendency to analyze this problem theoretically and derive an optimal U/P assignment beneath assumptions about signal model and filters. Furthermore, we tend to prove that the best U/P partition is connected to the correlation between nodes on the graph and is not the one that minimizes the number of conflicts (connections between nodes of same label) or maximizes the burden of the cut. We also provide experimental ends up in randomly generated graph signals and real information from image and video signals that validate our theoretical conclusions, demonstrating improved performance over state-of-the-art solutions for this problem.