Histograms of Oriented Gradients for Human Detection
CVPR '05 Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition (CVPR'05) - Volume 1 - Volume 01
Covariance Tracking using Model Update Based on Lie Algebra
CVPR '06 Proceedings of the 2006 IEEE Computer Society Conference on Computer Vision and Pattern Recognition - Volume 1
Steepest Descent For Efficient Covariance Tracking
WMVC '08 Proceedings of the 2008 IEEE Workshop on Motion and video Computing
Region covariance: a fast descriptor for detection and classification
ECCV'06 Proceedings of the 9th European conference on Computer Vision - Volume Part II
| Hi-index | 0.00 |
This paper presents a novel model, called orientation covariance matrices, to represent the object region and introduces a steepest descent method for object tracking. This model partitions the gradient orientation space of joint color channels into subspaces (bins), and computes the covariance matrix of image features in every bin. In the model a feature point does not belong exclusively to one bin; instead, it makes contributions to several neighboring bins. This is accomplished by introducing the cosine function for weighting the gradient components of feature vectors. The weighting function helps to alleviate the effect of errors in the computation of gradients induced by noise and illumination change. We also introduce a spatial kernel for emphasizing the feature vectors which are nearer to the object center and for excluding more background information. Based on the orientation covariance matrices, we introduce a distance metric and develop a steepest descent algorithm for object tracking. Experiments show that the proposed method has better performance than the traditional covariance tracking method.