Standards conversion using motion compensation

Abstract

For converting between television or high-definition television signals, which differ in line and field rates, improved standards converters have to be developed. After a short description of the general multi-dimensional sampling rate conversion, the degradations inherent in conventional converters are explained. The most disturbing artefacts are motion judder, motion blurring and vertical detail flashing. To improve the conversion, a motion-compensatinginterpolation filter is used in the new converter. To adapt the filter a motion vector estimation and a segmentation are performed. Global and local motion vectors are estimated using a block matching method. Incorrectly estimated vectors are eliminated by nonlinear postprocessing. The segmentation distinguishes between unchanged and changed parts of the image. Furthermore a suitable motion vector is defined for each pixel in the intermediate output field by selecting one of the surrounding block vectors or the global vector. Postprocessing of the pixel vectors is performed to eliminate incorrectly selected pixels vectors. Conventional standards converters and the improved converter were simulated on a computer.