Title :
On the Design of Optimal Sub-Pixel Motion Compensation Interpolation Filters for Video Compression
Author :
Minoo, Koohyar ; Baylon, David
Author_Institution :
ARRIS Group, San Diego, CA, USA
Abstract :
In this paper, the design of optimal temporal prediction for video coding is addressed as a quantization design problem. In the proposed framework, a codebook consisting of a set of interpolation filters is optimized to achieve rate-distortion optimality. The optimization process jointly affects two aspects of motion compensation to achieve rate distortion optimality: 1) The size of the codebook or motion vector (MV) resolution and 2) The filter coefficients for each sub-sample interpolation filter. Note that filter coefficients dictate the behavior of the interpolation filter in terms of signal-noise shaping.
Keywords :
data compression; image filtering; image resolution; interpolation; motion compensation; optimisation; prediction theory; quantisation (signal); video coding; codebook MV resolution; motion vector resolution; optimal subpixel motion compensation interpolation filter design; optimal temporal prediction design; optimization process; quantization design problem; rate distortion optimality; signal-noise shaping; subsample interpolation filter coefficient; video coding; video compression; Data compression; Finite impulse response filters; Interpolation; Motion compensation; Quantization (signal); Rate-distortion; Video compression; interpolation; motion compensation; uniform pase; video compression;
Conference_Titel :
Data Compression Conference (DCC), 2015
Conference_Location :
Snowbird, UT
DOI :
10.1109/DCC.2015.83