Add a search that will perform a hyper-localized search. This search effectively uses a small neighborhood around a given trajectory (both in terms of starting pixel and velocities) and returns all results from this neighborhood. This localized set can be used to:
refine trajectories by searching a finer parameter space around the best results found by the initial search, or
collect a distribution of trajectories and their likelihoods around a single result.
As part of this change we remove a small optimization from the core search (which did not bother to sort trajectories with LH <= -1.0 since we knew they would be filtered eventually). This removes that optimization so we can get the full grid of nearby trajectories.
Add a search that will perform a hyper-localized search. This search effectively uses a small neighborhood around a given trajectory (both in terms of starting pixel and velocities) and returns all results from this neighborhood. This localized set can be used to:
As part of this change we remove a small optimization from the core search (which did not bother to sort trajectories with LH <= -1.0 since we knew they would be filtered eventually). This removes that optimization so we can get the full grid of nearby trajectories.