Compressed Video Enhancement using Regularized Iterative Restoration Algorithms

Compressed Video Enhancement using Regularized Iterative Restoration Algorithms Passant Vatsalya Karunaratne Many applications, ranging from scienti c applications to consumer electronics products, use the acquisition and processing of digital video. The acquisition, compression, transmission, and format and rate conversion of an information carrying video signal degrades the accurate representation of the three dimensional continuous world conveyed by the original video signal. The visual quality of a video sequence, in the case of a human observer, and the information content of the video frames, in the case of image analysis, can be improved by methods of restoration of video sequences. Various video processing and restoration tasks can be performed by utilizing the additional information conveyed by the temporal dynamics of an image sequence, in addition to those possible with the application of single frame restoration techniques which operate on individual frames of an image sequence. This thesis addresses methods for video enhancement which make use of the information conveyed by the temporal dimension of dynamic image sequences, when the capturing, processing, transmission, and storage devices are unable to handle the required amount of information. The degradation of a video signal due to quantization is addressed in this work, and restoration methods for the reduction of compression artifacts, introduced by quantization and other processes associated with video compression, are presented.