An Integrated Approach to Grouping and Matching
Lists
Only you see this
Abstract
Perceptual grouping and relational matching are conventionally viewed as sequential stages of effective intermediate level scene interpretation. Relational structures established on the basis of perceptual grouping criteria are utilised in a bottom-up control strategy and hence form the input data representation for subsequent matching. Our standpoint in this paper is that the two processes should be tightly coupled to one-another so that the relational model can prevail upon the extraction of perceptual groupings, providing additional constraints on an otherwise potentially fragile processing operation. We realise this objective by casting the grouping and matching operations as an iterative discrete relaxation process. The dual operations of re-organising the perceptual relation graph and subsequent matching against a model, both optimise a single objective function in the maximum a posteriori probability sense. Grouping and matching are therefore cast into an integrated optimisation framework that is realised by tightly coupled update processes.
BibTeX
@inproceedings{wilson1995integrated,
author = {Richard C. Wilson and Edwin R. Hancock},
title = {{An Integrated Approach to Grouping and Matching}},
booktitle = {ICIAP},
year = {1995},
doi = {10.1007/3-540-60298-4\_237},
}