Output details
11 - Computer Science and Informatics
University of Dundee
Parts-based segmentation with overlapping part models using Markov chain Monte Carlo
<23> This paper appeared in a Special Issue of the journal based on expanded versions of selected papers from the British Machine Vision Conference. The conference acceptance rate was 30%. Based on the referees’ and area chairs’ recommendations as well as further reviews, 7% of the accepted papers were subsequently selected for the Special Issue. The paper presents a novel likelihood model for multiple overlapping parts and their inference using MCMC. It reports experiments, in collaboration with an orthopaedic surgeon, in which the method inferred bone boundaries in x-ray images as used for analysis of the progression of osteoarthritis.