AN EFFICIENT SEGMENTATION AND ACTION RECOGNITION IN VIDEO USING VISUAL SALIENCY AND RELIABLE REGION APPROACH

L.Sankari; S.Rekha

Volume : IV, Issue : VII, July - 2015

AN EFFICIENT SEGMENTATION AND ACTION RECOGNITION IN VIDEO USING VISUAL SALIENCY AND RELIABLE REGION APPROACH

L. Sankari, S. Rekha

Abstract :

A novel method for video segmentation is proposed for detection of small sized classes and to reduce the computational burden of algorithms. Maximum symmetric model approach and Exponential ellipse model are integrated to obtain the saliency value for each pixel. Secondly, the Gaussian Mixture Model (GMM) is applied to locate the object region. Visual saliency method and GMM method successfully segments the object. Then computational burden of the work can be reduced by exploring the idea of finding and using only the regions those are reliable for detection. To reduce the computational burden, object class segmentation is done with reliable regions. In addition, actions are recognized for the input images. First the color moment features are extracted from both training set and test data. Then the obtained features are compared with the training set.