Hidden Markov Model Based Structuring of Cricket Video Sequences Using Motion and Color Features.
Maheshkumar H. Kolekar, Somnath Sengupta · Indian Conference on Computer Vision, Graphics and Image Processing · 2004
In this paper, we propose our techniques and results on automatic analysis of cricket video to facilitate highlight generation and content-based retrieval. We use Dynamic Programming based on Hidden Markov Model (HMM-DP) approach for structure analysis of cricket video sequences. We introduce a scaling factor α with the likelihood values to achieve classification more robust and accurate. We obtained 87.5 % average overall classification for motion and 75 % for color, which clearly indicate that motion is more effective likelihood function than color for ‘Cricket’ video sequences. We classified real ‘Cricket’ video sequence into four semantically meaningful classes and demonstrated the validity of our approach.