Method of motion estimation for a digital input video signal

摘要

For motion estimation block matching can be used. The candidate motion vector to which the lowest matching error is related is selected as the motion vector. Subsampling of columns and/or lines can be used to reduce the computational and memory requirements. However, interlace signals require specific considerations because they are already vertically subsampled and a further vertical subsampling applied in the motion estimation processing can deteriorate severely the motion vector quality. Therefore the input signal (IP) is horizontally and vertically lowpass filtered (HVF) and horizontally subsampled (HSUB) by a factor of two. The motion vector search operates on the resulting video signal with a horizontal search step size of two (original) pixels and with a vertical search step size of one (original) pixel, whereby for calculating the cost function or block matching error in the reference block every second (original) pixel and (original) line only is evaluated. In an optional second block matching stage a fine motion estimation is carried out based on horizontally and vertically half-pel distance interpolated quincunx pixel patterns.