Open Collections

Using unitary transforms for MPEG block motion estimation

University of British Columbia

Block motion estimation, a stage of the MPEG video encoding process which significantly effects the quality and bit rate of an encoded video sequence, is also the most computation intensive part of the encoding process. To reduce the number of operations in this step, it is proposed that unitary transforms may be used to create condensed versions of the data, called block signatures, which can be used in place of the original video data during the matching process. Continuing previous work, this thesis studies several unitary transforms for their utility in creating signatures with the desired properties, with the focus being on how to select and allocate bits to the transform coefficients, and it considers the efficient computation of these signatures. Hardware and software implementations incorporating these signatures are suggested and discussed, and the results of implementing two proposed signature matching algorithms are described. A signature composed of sixty-four discrete cosine transform coefficients was found to work well in a full-search matching scheme appropriate for hardware implementation. The other proposed signature matching algorithm, which attempts to minimise the difference between the coefficients one at a time during the search for matches, performed poorly since the transform coefficients with most of the energy in the signal also tend contain most of the energy in the prediction error as well, and therefore the differences between these higher energy coefficients can not be minimised too early in the search process. Further work is needed to determine if a joint minimisation of differences between the transform coefficients of a target block and the set of available matches would be effective.

Thesis/Dissertation

application/pdf

2009-05-22

For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use.

http://hdl.handle.net/2429/8064

Electrical and Computer Engineering

University of British Columbia

UBCV

DSpace