Jump to navigation
Development of a CMSE based prediction model for HEVC video slices
The advancement of video technology and popularity of real time video applications in recent years have led to greater bandwidth requirement in wireless transmission. High efficiency video coding (HEVC) achieves the highest compression ratio so far. But when transmitted over the error-prone wireless channel, the quality of a compressed video is far more likely to get degraded with even a single lost packet, which leads to the need of error resiliency schemes. But because of the bandwidth limitation in networks, not all data can be protected equally. In this thesis, we develop a prediction scheme that would determine the distortion that the loss of a slice can cause to an HEVC encoded video. We utilize this knowledge to find the relative importance of the slices in the data stream to incorporate an unequal error protection scheme onto it. We select the cumulative mean squared error (CMSE) as the measure of distortion. But, as the actual measuring of CMSE involves summing up all the distortion introduced by a slice loss, starting from the current frame to the end of that group of picture (GOP), it is not suitable for real time video streaming applications. Therefore, instead of measuring the CMSE, we develop a low-complexity model to predict it. This makes the whole scheme suitable for real time applications. We generate a database consisting of several videos’ measured slice CMSE and other video factors extracted while encoding the videos. This database is then divided into a training set and a test set. Our prediction model is trained over the training set outputting a fitted model from it. Then the video factors of the test set are given as input to the fitted model to perform the CMSE prediction. After the CMSE is successfully predicted by the model, we propose several priority assignment schemes based on their predicted CMSE values and compare their results with each other. Finally these priority assigned slices are used to design an unequal error protection (UEP) scheme with forward error correction (FEC) codes.
Electrical and Computer Engineering
Master Of Science (M.S.) San Diego State University, 2017
© 2015 SDSU Library & Information Access. All Rights Reserved.