Optical flow by multi-scale annotated keypoints: A biological approach

• Author(s): Farrajota, Miguel ; Rodrigues, J. M. F. ; du Buf, J. M. H.
• Date: 2011
• Persistent ID: http://hdl.handle.net/10400.1/891
• Origin: Sapientia - Universidade do Algarve
• Subject(s): Visão humana
• Document Type: Conference Object
• Language: English

Description

Optical flow is the pattern of apparent motion of objects in a visual scene and the relative motion, or egomotion, of the observer in the scene. In this paper we present a new cortical model for optical flow. This model is based on simple, complex and end-stopped cells. Responses of end-stopped cells serve to detect keypoints and those of simple cells are used to detect orientations of underlying structures and to classify the junction type. By combining a hierarchical, multi-scale tree structure and saliency maps, moving objects can be segregated, their movement can be obtained, and they can be tracked over time. We also show that optical flow at coarse scales suffices to determine egomotion. The model is discussed in the context of an integrated cortical architecture which includes disparity in stereo vision.