Video applications are increasingly becoming popular in wired and wireless networks. On the other hand, large-scale digital video library systems are growing rapidly with the fast advancement in high-capacity storage devices. This would increase congestion in wireless networks due to their limited bandwidth. The congestion is dependent on the encoded bit rate for these video bitstreams. To solve this problem, video content based classification, which can be helpful in determining the appropriate video sending bit rate has received some attention recently. In the first part of the thesis we study the metrics; change in blurriness, change in brightness and sum of absolute differences that relate to the motion of the video and implemented a technique to classify videos based on their motion - slow, medium and fast motion. Classification is done based on two schemes cluster analysis and neural networks. Cluster analysis is a data analysis tool, which aims at sorting different objects into group if the degree of association between them is maximum and neural network is an information processing paradigm that is inspired by the way biological system, such as brain process information. The performances for clustering and artificial neural network schemes are measured using cophenetic correlation and MSE, respectively. We have achieved the cophenetic correlation of 86.7% and MSE of 1.64x10__. In the second part of thesis, we implemented an object-tracking algorithm using covariance matrix. Object tracking involves detecting of region of interest in every frame, which is very useful in region-of-interest aware video encoding. For this, we implemented a picture parameter set updating technique, which helps in updating the region of interest for every frame for H.264/AVC encoding. The covariance-tracking algorithm performs well when the object undergoes both rigid and non-rigid motion..