Interactive Evolutionary Parameter Optimization for Eulerian Video Magnification
Syed Raazi Rizvi, Shahryar Rahnamayan · 2018
This paper explores control parameter tuning for the Eulerian Video Magnification (EVM) process using a pair comparison-based interactive differential evolution (PCB-IDE) algorithm. Interactive evolutionary optimizers have a history of being applied to video/image post-processing tasks because assessing the quality of video data is objectively difficult without subjective assessment by users. The EVM technique magnifies motion and color variations within a video in a manner similar to a visual microscope. Applying EVM to video sequences allows the observation of known physical phenomenon by using only video data, which is critical to development of camera based monitoring applications. The proposed PCB-IDE assists experts in determining the optimal parameters for EVM process to better identify visual changes and correlate them to physical phenomenons.