Motion Vector Refinement for MPEG-2 to H.264 Transcoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video transcoding methods between MPEG-2 and H.264 formats are inefficient due to differences in coding modes, motion compensation, and hardware architectures, leading to sub-optimal coding efficiency and unnecessary complexity.
Innovation Solution
A method for transcoding that refines motion vectors to quarter-pel accuracy, reduces partition sizes, and utilizes MPEG-2 coding mode information in H.264 encoding, by generating a decoded frame, determining a search center, and searching for refined motion vectors in a temporal direction to improve motion compensation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If conventional transcoding methods are used that simply decode in MPEG-2 and re-encode in H.264, then the transcoding process is simple to implement, but coding efficiency is sub-optimal and complexity reduction methods are unnecessary for modern hardware architectures
Solution Approach 1:
The patent performs motion estimation refinement in advance during the transcoding process. Instead of directly re-encoding MPEG-2 decoded data into H.264, the method预先 conducts motion estimation on the decoded macroblocks to generate refined motion vectors, which are then used to seed the H.264 encoding process. This preliminary action improves coding efficiency by providing better motion compensation while maintaining reasonable implementation complexity.
2Device complexity
If MPEG-2 motion vectors are reused without refinement, then the transcoding complexity is reduced, but quarter-pixel accurate motion compensation cannot be achieved and coding efficiency suffers
Solution Approach 1:
The patent changes the precision parameter of motion vectors from half-pixel (MPEG-2 standard) to quarter-pixel (H.264 standard). This is achieved by performing motion estimation refinement that searches at quarter-pel positions using the MPEG-2 decoded macroblocks as reference. The refined quarter-pixel motion vectors are then used for H.264 motion compensation, improving prediction accuracy while managing complexity through selective refinement.
Solution Approach 2:
The patent extends the motion vector search from the traditional half-pixel grid to a finer quarter-pixel grid, effectively adding a dimension of precision. This dimensional change allows for more accurate motion compensation by searching at positions between the coarser half-pixel locations, thereby improving coding efficiency without requiring complete re-encoding.
3Productivity
If motion estimation refinement is performed for all possible H.264 motion compensation partitions, then coding efficiency is optimized, but the transcoding complexity increases significantly
Solution Approach 1:
The patent applies motion estimation refinement selectively based on local characteristics of each macroblock. Instead of uniformly refining all macroblocks, the method identifies regions that benefit most from refinement (such as areas with significant motion or complex textures) and applies the computationally intensive quarter-pixel search only to those regions. This local quality approach optimizes coding efficiency where needed while controlling overall complexity.
Solution Approach 2:
The patent performs partial motion estimation refinement rather than complete refinement for all macroblocks and all partition sizes. The method selectively applies refinement to certain macroblocks or certain partition sizes based on complexity constraints, achieving a balance between coding efficiency improvement and computational complexity management. This partial action approach provides sufficient coding gain without the full burden of exhaustive refinement.
4Productivity
If MPEG-2 field/frame macroblock decisions are reused independently for each macroblock, then the transcoding process is efficient, but H.264 macroblock pair coding cannot be properly utilized
Solution Approach 1:
The patent creates a universal macroblock representation that works for both MPEG-2 and H.264 coding structures. By decoding MPEG-2 macroblocks and performing motion estimation to generate refined motion vectors, the method creates a unified processing framework that can then adapt to H.264's macroblock pair structure. This universal approach allows the same core processing logic to serve both standards while maintaining their specific coding requirements.
Solution Approach 2:
The patent dynamically adapts the coding structure from MPEG-2's independent macroblock decisions to H.264's macroblock pair structure. The method uses MPEG-2 decoded data as input but dynamically reorganizes and processes the data according to H.264's MBAFF (macroblock adaptive field/frame) structure. This dynamic adaptation allows proper utilization of H.264 coding modes while efficiently processing the incoming MPEG-2 stream.
Data Source
AI summary
A method for transcoding that includes (A) generating a decoded frame by decoding an MPEG-2 input video stream, the decoded frame including decoded macroblocks; (B) determining a search center for encoding a current macroblock corresponding to a pair of the decoded macroblocks on consecutive macroblock rows, when (i) the encoding uses a predictive field mode and (ii) a current field being encoded is a second field of a current frame that has a first field as a reference field, the search center is a temporally scaled version of a decoded motion vector from an upper or a lower macroblock in the pair; (C) generating a refined motion vector by searching in a temporal search direction about the search center; and (D) generating an H.264 output video stream by encoding the current macroblock based on the refined motion vector.