A Bi-Directional Integrated Model for Non-Rigid Motion Analysis

To be able to understand the motion of deformable objects, techniques in image processing and computer vision are essential for non-rigid motion analysis in this active research area. We have developed an integrated model that combines the advantages of the boundary-basedand region-based approaches and avoids problems caused by each stand-alone approach, e.g. overshrinking, oversegmenting, noise sensitivity. This image segmentation model further iteratively improves each submodel in both directions until it satisfies predefined criteria. Different frame-to-frame prediction methods, naive, inflation, and optic flow, are developed and evaluated. Comparison between our model and other models is studied and illustrated by examples. Further improvements on our motion tracking model are possible by evaluating new attraction functions and prediction methods. Some important notes on future work are given for incorporating other information (e.g. morphometrics) into our integrated model in sev...

• Type: Conference
• Year: 1998
• Where: BMVC
• Authors: Chunyen Liu
• Keywords: BMVC 1998, Computer Vision, Integrated Model, Model, Motion Analysis