Motion vector refinement for MPEG-2 to H.264 video transcoding

A system comprising:

• a decoder device configured to generate a decoded frame by decoding an input video stream in an MPEG-2 format, said decoded frame including a plurality of decoded macroblocks;
• and
• an encoder configured to (i) generate a current macroblock in a current field by transferring a plurality of lines from a pair of said decoded macroblocks to said current macroblock, wherein (a) said pair of decoded macroblocks are on consecutive macroblock rows in said decoded frame and (b) said current macroblock receives half of said lines from each of said decoded macroblocks in said pair, (ii) select a decoded motion vector from one of an upper macroblock or a lower macroblock of said pair in response to (a) an encoding of said current field using a predictive field mode, (b) said current field being a second field of a current frame and (c) said current field using a first field of said current frame as a reference field in said predictive field mode, (iii) generate an intermediate motion vector by temporally scaling said selected decoded motion vector to point to said first field of said current frame, (iv) generate a current motion vector by performing an H.264 motion vector search in a temporal search direction within an area having a center determined by said intermediate motion vector and (v) generate an output video stream in an H.264 format by said encoding of said current macroblock based on said current motion vector, wherein said decoded motion vector is selected from said upper macroblock in response to (i) said upper macroblock having at least one motion vector matching said temporal search direction, (ii) said lower macroblock having no motion vectors matching said temporal search direction, (iii) said upper macroblock having a field motion compensation and (iv) said decoded motion vector matching a polarity of said current field.