Towards a Robust Spatio-Temporal Interest Point Detection for Human Action Recognition

Spatio-temporal salient features are widely being usedfor compact representation of objects and motions in video,especially for event and action recognition. The existingfeature extraction methods have two main problems: First,they work in batch mode and mostly use Gaussian (linear)scale-space filtering for multi-scale feature extraction. This linear filtering causes the blurring of the edges and salient motions which should be preserved for robust eature extraction.Second, the environmental motion and ego disturbances(e.g., camera shake) are not usually differentiated.These problems result in the detection of false features nomatter which saliency criteria is used. To address theseproblems, we developed a non-linear (scale-space) filteringapproach which prevents both spatial and temporal dislocations.This model can provide a non-linear counterpart ofthe Laplacian of Gaussian to form the conceptual structuremaps from which multi-scale spatio-temporal salient features are extracted. Preliminary evaluation shows promising result with false detection being removed.