Abstract: One embodiment of an edge-preserving vertical interpolation system constructs a de-interlaced video frame on a pixel-by-pixel basis using an edge-preserving vertical interpolation technique. Pixels within a pixel window centered about a selected pixel determine the direction of an intensity gradient associated with the selected pixel. A first pixel is constructed by interpolating between pixels that are perpendicular to the direction of the intensity gradient and a confidence factor is computed that indicates the likelihood that there is only one edge within the pixel window. A second pixel is constructed using a non-edge-preserving vertical interpolation technique. Interpolating between the first pixel and the second pixel based on the confidence factor generates a pixel in the de-interlaced video frame corresponding to the selected pixel.

Abstract: One embodiment of a motion-adaptive video de-interlacing system includes a motion estimator and a pixel interpolator. The motion estimator determines the magnitude of motion associated with each pixel within a de-interlaced video frame. In some instances, as determined by the motion values, the pixel interpolator produces final pixel values by blending between pixel values produced through different de-interlacing methods optimized for different levels of pixel motion. The present invention advantageously produces de-interlaced video frames having relatively better picture quality than those produced using prior art techniques, especially for small pixel motions.