Show simple item record

dc.rights.licenseRestricted to current Rensselaer faculty, staff and students. Access inquiries may be directed to the Rensselaer Libraries.
dc.contributorWoods, John W. (John William), 1943-
dc.contributorKar, Koushik
dc.contributorTajer, Ali
dc.contributorKovacic, Gregor
dc.contributor.authorLiu, Yuan
dc.date.accessioned2021-11-03T09:16:08Z
dc.date.available2021-11-03T09:16:08Z
dc.date.created2020-08-10T12:02:39Z
dc.date.issued2019-12
dc.identifier.urihttps://hdl.handle.net/20.500.13015/2500
dc.descriptionDecember 2019
dc.descriptionSchool of Engineering
dc.description.abstractFinally, we investigated the selection of Lagrange multipliers for motion estimation based on extracted features of video clips to be compressed by interframe EZBC. By designing an iterative mechanism, we showed the PSNR performance of decoded video clips at certain bitrate can be greatly affected by modifying the Lagrange multipliers with respect to the clip or even multiple segments within a video sequence. Parameters related to the feature of video clip segments are extracted to show where and how Lagrange multipliers of each temporal level should be adaptively reselected. Experimental results regarding both classic test video clips as well as the artificial concatenations of them are obtained and discussed, which is then followed by building a theoretical model of Lagrange multiplier versus target bitrate.
dc.description.abstractThen we explored two-component motion compensation modes, which attempt to detect and encode two distinct moving objects in a single block. By carefully investigating related research and techniques, we developed a new motion compensation mode that utilizes spatial texture and inter-frame information to detect two distinct motion patterns within a single block for motion compensation. With the help of a Canny edge detector, the MCTF process for two-component blocks can be achieved. Unfortunately, the method in its present form based on Interframe EZBC has proved to be only beneficial on a synthesized test clip. Further work is thus suggested to improve this approach under the highly-scalable video coding environment.
dc.description.abstractNowadays, the development of video coders results in significantly increasing performance. Among all categories, highly-scalable video coders have a chance to show a great storage advantage by freely extracting bitstreams with different bitrate, resolution, and frame-rates from a master bitstream according to the various requirements of heterogeneous subscribers in a multicast network. Based on the enhanced MC-EZBC scalable video coder with temporal decomposition structure and multi-mode motion-compensation system, we revise and extend the motion-compensation temporal filtering (MCTF) framework to incorporate advanced mechanisms such as motion-vector predictor competition, block-merging, and affine motion-compensation modes. The upper limit of macro motion block size is also increased to improve the efficiency of motion estimation. Besides, we also provide an alternative backend based on JPEG 2000 to facilitate multi-level code-stream extraction. Experimental results showed their ability to both significantly improve the PSNR performance and also extend the scenario of scalable video coding. By modifying the MCTF framework, we designed and created the innovative coders named Interframe EZBC and Inteframe JPEG 2000.
dc.language.isoENG
dc.publisherRensselaer Polytechnic Institute, Troy, NY
dc.relation.ispartofRensselaer Theses and Dissertations Online Collection
dc.subjectElectrical engineering
dc.titleAn improved video compression framework with fine-grain scalability
dc.typeElectronic thesis
dc.typeThesis
dc.digitool.pid179975
dc.digitool.pid179976
dc.digitool.pid179977
dc.rights.holderThis electronic version is a licensed copy owned by Rensselaer Polytechnic Institute, Troy, NY. Copyright of original work retained by author.
dc.description.degreePhD
dc.relation.departmentDept. of Electrical, Computer, and Systems Engineering


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record