Nova Patents
EP1276331A2

Method for preventing dual-step half-pixel motion compensation accumulation errors in prediction-rich MPEG-2 sequences

Abstract

A two-step motion prediction for MPEG-2 interpolation case-D will yield visual artifacts if not corrected. An improved MPEG-2 decoder includes a logic gate, multiplexer, and adder. When both the horizontal (h0) and vertical (h1) motion vector components require a half pixel interpolation (case-D), the multiplexer forwards the constant minus three to the adder, otherwise a constant zero is used. Such adder modifies the DC coefficient input to the inverse discrete cosine transformer to include a correction term for the predicted pixels calculated by a two-step predictor. A correction value of -0.375 is evenly distributed over all sixty-four resulting spatial coefficients during the inverse discrete cosine transform. This results statistically in a slightly brighter set of correction terms. Such offsets result in a slightly darker prediction that is formed by the two-step predictor. The output frames are statistically correct images.

EP1276331A2, drawing sheet 1
Sheet 1 of 86

Term

Term ended

Projected expiry passed 31 March 2020, 6.5 years ago.

  1. Priority
  2. Filed
  3. Published
  4. Projected expiry
  5. Today

8 claims: 2 independent, 6 dependent

  1. 1
    A method for decompressing MPEG-2 datastreams, comprising:separating an MPEG-2 datastream into a motion vector datastream and a coefficients datastream;using a two-step half-pixel prediction process on said motion vector datastream that produces a rounding error when a predicted frame is used as a reference frame in a subsequent prediction;and modifying said coefficients datastream before entering an inverse discrete cosine transform stage in a way that results in a zeroing out of pumping artifacts in a frame store.
  2. 6
    An MPEG-2 decoder, comprising:a header parser for separating an MPEG-2 datastream into a motion vector datastream and a coefficients datastream;a motion vector calculator connected to receive said motion vector datastream and for providing a vertical motion vector component h 1 and a horizontal motion vector component h 0 ;a coefficients decoder connected to receive said coefficients datastream and for deriving a DC coefficient and a plurality of AC coefficients with a variable length decoder, a run length decoder, and an inverse quantization unit;a two-step half-pixel prediction processor connected to receive a motion vector output from the motion vector calculator, and for providing a predicted frame with a rounding error that occurs when said predicted frame is used as a reference frame in a subsequent prediction;a logic unit connected to receive said DC coefficient and said vertical motion vector component h 1 and a horizontal motion vector component h 0 , and providing for a modification of said DC coefficient to produce a modified DC coefficient that is dependent on said vertical motion vector component h 1 and said horizontal motion vector component h 0 ;an inverse discrete cosine transformer connected to receive said modified D C coefficient from the logic unit and plurality of AC coefficients from the coefficients decoder;and a combiner for summing an output of the two-step half-pixel prediction processor and the inverse discrete cosine transformer, and for providing a series of frames;wherein, said DC coefficient is modified before entering the inverse discrete cosine transformer in a way that results in a zeroing out of pumping artifacts in said frames that would otherwise be caused by rounding errors that occur in the two-step half-pixel prediction processor.
Independent claims2