Multiview video compression with 1-D transforms

Download
2013
Karasoy, Burcu
In previous research, it has been shown that motion compensated prediction residuals can have 1-D structures in many regions and that 1-D directional DCTs can compress such regions more e ciently than the conventional 2-D DCT. In this thesis, we analyze the spatial characteristics of the disparity compensated prediction residuals and the analysis results show that, similar to motion compensated prediction residuals, many regions of disparity compensated prediction residuals also have 1-D structures. Thus, we apply 1-D directional DCTs for compressing disparity compensated prediction residuals in this thesis. To show the achievable compression e ciency gains from using these transforms, we modify the reference software of the Multiview Video Coding (MVC) extension of H.264 so that each residual block can be transformed either with one of the 1-D DCTs or the conventional 2-D DCT.Experimental results indicate that the overall compression e ciency increases.
Citation Formats
B. Karasoy, “Multiview video compression with 1-D transforms,” M.S. - Master of Science, Middle East Technical University, 2013.