Low bandwidth, two-candidate motion estimation for interlaced video

A method for performing motion estimation on a sequence of video pictures comprising the steps of:

• (a) processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• (b) using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively, and
• (c) using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• wherein said reference picture recedes a to-be-encoded picture, said method further comprising the steps of:
• (d) processing a second reference picture that succeeds said to-be-encoded picture to generate fourth, fifth and sixth pictures, said fourth picture having a lower resolution than said fifth picture and said fifth picture having a lower resolution than said fourth picture,
• (e) performing first and second stage field motion estimation searches using said fourth and fifth pictures, respectively, to obtain a fourth field motion vector candidate and a fifth field motion vector candidate in said first and second stages, respectively, and
• (f) using at least one of said fourth and fifth field motion vector candidates as a starting position and said sixth picture, performing a third stage field motion estimation search.

A method for performing motion estimation on a sequence of video pictures comprising the steps of:

• (a) processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• (b) using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively,
• (c) using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• (d) performing frame motion estimation on said reference picture, and
• (e) determining whether to predict each of said macroblocks of said to-be-encoded frames as a frame macroblock using prediction information derived from said frame motion estimation or as plural field macroblocks using prediction information derived from said third stage field motion estimation searches.

A method for performing motion estimation on a sequence of video pictures comprising the steps of:

• (a) processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• (b) using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively,
• (c) using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• (d) performing said third stage field motion estimation search using said first field motion vector candidate as a starting position and said third picture to produce a third motion vector and performing said third stage field motion estimation search using said second field motion vector candidate as a starting position and said third picture to produce a fourth motion vector,
• (e) selecting only one of said third and fourth motion vectors, and
• (f) performing a fourth stage motion estimation search in said reference picture using said selected motion vector as an initial starting position, said third picture having a lower resolution than said reference picture.

A method for performing motion estimation on a sequence of video pictures comprising the steps of:

• (a) processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• (b) using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively,
• (c) using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• wherein said reference picture is a frame and said first, second and third pictures are scaled versions of a first field of said reference picture, said method further comprising the steps of:
• (d) processing said reference picture to produce fourth, fifth and sixth fields, wherein said fourth, fifth and sixth fields are scaled versions of a second field of said reference picture, said fourth field having a lower resolution than said fifth field and said fifth field having a lower resolution than said sixth field,
• (e) performing first and second stage field motion estimation searches using said fourth and fifth fields, respectively, to obtain a fourth field motion vector candidate and a fifth field motion vector candidate in said first and second stages, respectively,
• (f) using at least one of said fourth and fifth field motion vector candidates as a starting position and said sixth field, performing a third stage field motion estimation search to obtain a sixth field motion vector candidate,
• (g) selecting either said third field motion vector candidate and said first field of said reference picture or said sixth field motion vector candidate and said second field of said reference picture, and
• (h) using said selected field motion vector candidate as a starting position and said selected field of said reference picture, performing a fourth stage field motion estimation search.

A method for performing motion estimation on a sequence of video pictures comprising the steps of:

• (a) processing a reference picture to generate, first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• (b) performing a first stage field motion estimation search using said first picture to obtain a first motion vector from a top reference field and a second motion vector from a bottom reference field,
• (c) performing a second field motion estimation search using said second picture to obtain a third motion vector from said top reference field and a fourth motion vector from said bottom reference field,
• (d) selecting one of said first and second motion vectors as a fifth motion vector and selecting one of said third and fourth motion vectors as a sixth motion vector, and
• (e) using said fifth and sixth motion vectors as starting positions and said third picture, performing a third stage field motion estimation search to generate seventh and eighth motion vectors, respectively.

An apparatus for performing motion estimation on a sequence of video pictures comprising:

• a preprocessor for processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture, and
• a motion estimator for, using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively, and for using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• wherein said reference picture precedes a to-be-encoded picture,
• wherein said preprocessor is also for processing a second reference picture that succeeds said to-be-encoded picture to generate fourth, fifth and sixth pictures, said fourth picture having a lower resolution than said fifth picture and said fifth picture having a lower resolution than said fourth picture, and
• wherein said motion estimator is also for performing first and second stage field motion estimation searches using said fourth and fifth pictures, respectively, to obtain a fourth field motion vector candidate and a fifth field motion vector candidate in said first and second stages, respectively, and for, using at least one of said fourth and fifth field motion vector candidates as a starting position and said sixth picture, performing a third stage field motion estimation search.

An apparatus for performing motion estimation on a sequence of video pictures comprising:

• a preprocessor for processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• a motion estimator for, using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively, and for using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• wherein said motion estimator is also for performing frame motion estimation of said reference picture, and
• wherein said motion estimator is also for determining whether to predict each of said macroblocks of said to-be-encoded frames as a frame macroblock using prediction information derived from said frame motion estimation or as plural field macroblocks using prediction information derived from said third stage field motion estimation searches.

An apparatus for performing motion estimation on a sequence of video pictures comprising:

• a preprocessor for processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• a motion estimator for, using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively, and for using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• wherein said motion estimator is also for performing said third stage field motion estimation search using said first field motion vector candidate as a starting position and said third picture to produce a third motion vector and performing said third stage field motion estimation search using said second field motion vector candidate as a starting position and said third picture to produce a fourth motion vector, for selecting only one of said third and fourth motion vectors, and for performing a fourth stage motion estimation search in said reference picture using said selected motion vector as an initial starting position, said third picture having a lower resolution than said reference picture.

An apparatus for performing motion estimation on a sequence of video pictures comprising:

• a preprocessor for processing a reference picture to generate first, second and third pictures, said first picture having a lower resolution than said second picture and said second picture having a lower resolution than said third picture,
• a motion estimator for, using independent starting positions, performing first and second stage field motion estimation searches using said first and second pictures, respectively, to obtain a first field motion vector candidate and a second field motion vector candidate in said first and second stages, respectively, and for using at least one of said first and second field motion vector candidates as a starting position and said third picture, performing a third stage field motion estimation search to obtain a third field motion vector candidate,
• wherein said preprocessor processes a frame as said reference picture, said first, second and third pictures being scaled versions of a first field of said reference picture, said preprocessor also for processing said reference picture to produce fourth, fifth and sixth fields, wherein said fourth, fifth and sixth fields are scaled versions of a second field of said reference picture, said fourth field having a lower resolution than said fifth field and said fifth field having a lower resolution than said sixth field, and
• wherein said motion estimator is also for performing first and second stage field motion estimation searches using said fourth and fifth fields, respectively, to obtain a fourth field motion vector candidate and a fifth field motion vector candidate in said first and second stages, respectively, for, using at least one of said fourth and fifth field motion vector candidates as a starting position and said sixth field, performing a third stage field motion estimation search to obtain a sixth field motion vector candidate, for selecting either said third field motion vector candidate and said first field of said reference picture, and for, using said selected field motion vector candidate as a starting position and said selected field of said reference picture, performing a fourth stage field motion estimation search.