作者: Matthew Shreve , Sridhar Godavarthy , Vasant Manohar , Dmitry Goldgof , Sudeep Sarkar
DOI: 10.1109/WACV.2009.5403044
关键词: Segmentation 、 Image segmentation 、 Optical flow 、 Spotting 、 Object detection 、 Computer science 、 Facial expression 、 Face (geometry) 、 Computer vision 、 Artificial intelligence 、 Expression (mathematics)
摘要: This paper presents a novel method for automatic spotting (temporal segmentation) of facial expressions in long videos comprising continuous and changing expressions. The utilizes the strain impacted on skin due to non-rigid motion caused during magnitude is calculated using central difference over robust dense optical flow field each subjects face. Testing has been done 2 datasets (which includes 100 macro-expressions) promising results have obtained. several common drawbacks found expression segmentation including moderate in-plane out-of-plane motion. Additionally, also modified work with containing micro-expressions. Micro-expressions are detected utilizing their smaller spatial temporal extent. A subject's face divided sub-regions (mouth, cheeks, forehead, eyes) these regions. Strain patterns individual regions used identify subtle changes which facilitate detection