% IMPORTANT: The following is UTF-8 encoded. This means that in the presence % of non-ASCII characters, it will not work with BibTeX 0.99 or older. % Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or % “biber”. @PHDTHESIS{Thakur:765565, author = {Thakur, Uday Singh}, othercontributors = {Ohm, Jens-Rainer and Bull, David}, title = {{P}erceptual video coding using steerable pyramids}, school = {Rheinisch-Westfälische Technische Hochschule Aachen}, type = {Dissertation}, address = {Aachen}, reportid = {RWTH-2019-07494}, pages = {1 Online-Ressource (xiv, 98 Seiten) : Illustrationen, Diagramme}, year = {2018}, note = {Veröffentlicht auf dem Publikationsserver der RWTH Aachen University 2019; Dissertation, Rheinisch-Westfälische Technische Hochschule Aachen, 2018}, abstract = {In the context of this thesis we are mainly focused on developing coding tools for static and dynamic textures. Both static and dynamic textures are challenging to code for the current generation video codecs. Former is characterized with high amount of spatial variations which are randomly distributed over the textured region. At low bitrates such a content often shows relatively higher blocking and blurring artefacts as compared to a non-textured content. As a proposed solution to this problem, texture synthesis based on steerable pyramids is used at the decoder side for generating artificial textures using coded parameters. The results provided in this thesis prove that a much better visual quality is achievable at low bitrates when compared to the HEVC. Dynamic textures are characterized with complex motion patterns e.g. water waves, leaves swirling in the wind etc. Due to highly dynamic nature of the motion in such a content, motion compensation often fails to find a good predictor and as a result B-pictures are expensive to code. Under low bit-rate conditions, heavy blocking and blurring artefacts are often observed in the B-pictures. In our proposed solution for the dynamic textures we target those pictures that are most susceptible to these artefacts (i.e. pictures corresponding to two highest temporalids) and code these pictures in a reduced resolution format. At the decoder side these pictures are up-sampled with high-frequency synthesis using information from the neighboring full resolution pictures. The results provide a concrete evidence that our model outperforms JEM at the same rate as our overall perceptual quality is better in majority cases when viewing dynamic textures.}, cin = {613210}, ddc = {621.3}, cid = {$I:(DE-82)613210_20140620$}, typ = {PUB:(DE-HGF)11}, doi = {10.18154/RWTH-2019-07494}, url = {https://publications.rwth-aachen.de/record/765565}, }