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.contributorRadke, Richard J., 1974-
dc.contributor.authorChen, Di
dc.date.accessioned2021-11-03T08:17:38Z
dc.date.available2021-11-03T08:17:38Z
dc.date.created2015-03-09T10:00:50Z
dc.date.issued2014-12
dc.identifier.urihttps://hdl.handle.net/20.500.13015/1277
dc.descriptionDecember 2014
dc.descriptionSchool of Engineering
dc.description.abstract(4) Legendre series is usually used in surface fitting and signal processing. In this thesis, it is used to improve several important steps in the motion compensation. The author first modified the Legendre series formulae to make them suitable for processing image signals in motion compensation. Then a certain Legendre coefficient based factor is proposed to modulate the Lagrange factor in the variable block-size division criteria. The block mode decision criterion is then proposed to be improved by a block-based residual frequency analysis using Legendre series. The thesis also devised a Legendre based corner detector that could be helpful to resist the noise and interference in the block splitting decisions. Unfortunately, time did not permit including these Legendre series based improvements into ENH-MC-EZBC, but only some preliminary results were obtained.
dc.description.abstract(5) The reduction of predicition residual is verified through calculation. For test sequences with large area of affine motion, the decoding outputs of the new program result in better compression performance in terms of both PSNR and visual appearance.
dc.description.abstractThe variable size block-matching method used in ENH-MC-EZBC is both simple and effective with low computational complexity. However, it fails to capture complex motions such as rotation, scaling, and other deformations of moving objects.
dc.description.abstractThe ENH-MC-EZBC is a noteworthy fully scalable video coding schemes that uses a motion-compensated temporal filter (MCTF) to utilize the interframe correlation. The subband/wavelet transform within EZBC is then used on the various temporal components to take advantage of spatial or intraframe correlation.
dc.description.abstractThis thesis uses the exisitng ENH-MC-EZBC as a basis. Then some adjustments are made in the motion estimation and compensation steps to make them more applicable for affine and bilinear motion:
dc.description.abstract(1) The new coder inherits the method of quadtree block division and the matching algorithm for searching block MVs. However, the method of calculating pixel MVs from block MVs is different. Here, the blocks are divided into sub-blocks where bilinear interpolation is used to estimate pixel MVs which are preferred for objects with more general non-translational motion. The MCTF is then calculated with the more general motion model.
dc.description.abstract(2) An adaptive method which combines conventional block matching and bilinear motion model is proposed. Because the bilinear motion model is preferred in areas which are inside the block while block matching is preferred on the edge of objects, the project compares the residual from both methods and chooses the one end up with smaller residual. The method ensures at least the same compression performance as that of conventional block matching if not improved. The extra bit required is negligible.
dc.description.abstract(3) This thesis develops an adjustment algorithm to make the block MVs more accurate for the bilinear motion model. In the motion estimation step, we still use the existing block matching algorithm to find block MVs. So the block MVs are optimized under the assumption that block matching motion compensation will be used afterwards not the bilinear-based motion model. The difference between the two reduces the accuracy of block MVs. So a small adjustment of block MVs is added to be more suitable for bilinear motion model.
dc.description.abstractDue to the increasing demand for transmitting video over channels with heterogenous bandwidth, a fully scalable video encoding scheme is needed to transmit compressed videos at various qualities and resolutions.
dc.language.isoENG
dc.publisherRensselaer Polytechnic Institute, Troy, NY
dc.relation.ispartofRensselaer Theses and Dissertations Online Collection
dc.subjectElectrical engineering
dc.titleNew bilinear model for inter-frame video compression
dc.typeElectronic thesis
dc.typeThesis
dc.digitool.pid174706
dc.digitool.pid174707
dc.digitool.pid174708
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.degreeMS
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