A dynamic block scale-invariant (DBSI) method for extracting salient feature key points that can be used to perform fast and accurate sparse optical flow computation is presented. With a more in-depth development and analysis of the Dynamic Block-Based feature selection and the Scale-space Extrema Detection stage, we propose a solution that correctly estimated the optical flow 2D motion vectors in the region of DBSI salient features, which are robust against noise, brightness and scale changes. The effectiveness of our method is demonstrated on the Middlebury optical flow benchmark and the experiment results show that it is competitive with the state-of-the-art.
Salient Feature Detection for optical flow computation
01.08.2016
242222 byte
Aufsatz (Konferenz)
Elektronische Ressource
Englisch
Salient Object Detection: A Discriminative Regional Feature Integration Approach
British Library Online Contents | 2017
|Selective salient feature based lane analysis
IEEE | 2013
|