Long-term memory motion compensated prediction and up to ¼-pel accurate motion compensation contribute a considerable portion of the compression gain provided by H.264/AVC over its predecessors. This paper investigates the factors contributing to the spectral distortions introduced in the digitized video signal. A quantitative analysis shows that fractional-pel interpolation is the main source of these spectral distortions. Using these results, two techniques are proposed for reducing computational complexity with negligible effects on the quality of the video. Simulation of the proposed techniques show up to 56% complexity reduction compared to the reference scheme without any significant decrease in signal-to-noise ratio.