Abstract
In this thesis, we propose a system to analyze and classify the video shots of the baseball game TV program into fifteen categories. Our system consists of three modules: feature extraction, Hidden Markov Model training, and video shot categorization. First, we analyze the motion, color, and texture information of input image sequence to generator our feature vector. Then, to train different HMMs, we use different training set of video shots. Finally, for an input video shot, we apply all the trained HMMs to find the most probably HMM and assign the corresponding category to the input video shot. The experimental results show that the average recognition rate is 84.72%.