You have requested a machine translation of selected content from our databases. This functionality is provided solely for your convenience and is in no way intended to replace human translation. Neither SPIE nor the owners and publishers of the content make, and they explicitly disclaim, any express or implied representations or warranties of any kind, including, without limitation, representations and warranties as to the functionality of the translation feature or the accuracy or completeness of the translations.
Translations are not retained in our system. Your use of this feature and the translations is subject to all use restrictions contained in the Terms and Conditions of Use of the SPIE website.
7 July 1998Multiresolution motion estimation and target detection
Detecting and characterizing motion in a scene can play a critical role in target detection algorithms, since many targets can be camouflaged so completely that, if they are not moving, they are nearly undetectable. However, once they begin moving, they `popout' and are immediately detected. Estimating motion is also important in human vision modeling, because motion is generally detected with peripheral vision, which can cover the field of regard much more quickly than foveal vision. In this paper, we present two hierarchical multiresolution methods for computing the optical flow in a scene. We use statistical properties of the resulting flow fields to compute a motion feature vector, which we relate to the conspicuity of the moving target in a scene via a neural network.
The alert did not successfully save. Please try again later.
Robert E. Karlsen, David J. Gorsich, Grant R. Gerhart, "Multiresolution motion estimation and target detection," Proc. SPIE 3375, Targets and Backgrounds: Characterization and Representation IV, (7 July 1998); https://doi.org/10.1117/12.327150