Motion Vector Selection Using Statistically Popular Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video encoding methods face challenges in achieving optimal motion vector selection, particularly when encoding frames with large objects translating across backgrounds, as they tend to prioritize local motion vector continuity over global optimality, leading to suboptimal compression and visual quality.
Innovation Solution
The method involves computing a rate distortion value for each test vector position using a linear combination of distortion and bit cost values, incorporating statistically popular motion vectors to account for expected future benefits, allowing for a more global motion vector selection that aligns with object boundaries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of substance
If conventional rate distortion optimization is used with large lambda values to minimize bit rate, then compression ratio is improved, but motion vector selection becomes locally optimal rather than globally optimal, leading to misalignment with true object motions
Solution Approach 1:
The patent performs preliminary action by computing a set of statistically popular motion vectors for the current frame before the main encoding process. These pre-computed vectors represent the most likely motion patterns in the frame and are stored for use during rate distortion optimization, enabling blocks to reference globally optimal motion vectors rather than relying solely on local prediction vectors.
Solution Approach 2:
The patent introduces an intermediary mechanism by adding a bias term to the rate distortion function that incorporates the distance between candidate motion vectors and the statistically popular motion vectors. This bias acts as a mediator that guides the optimization process toward globally optimal vectors while still allowing local variations when necessary, thus resolving the conflict between bit rate minimization and alignment accuracy.
2Stability of the object's composition
If motion vector search emphasizes local continuity through prediction vectors, then spatial inertia is maintained, but global motion patterns such as large object translations across backgrounds are not captured
Solution Approach 1:
The patent computes statistically popular motion vectors for the entire frame before processing individual blocks. These pre-computed global motion vectors serve as a reference that blocks can use to align with true object motions, enabling the system to adapt to global motion patterns while maintaining local continuity where appropriate.
Solution Approach 2:
The patent introduces a bias term as an intermediary that bridges local prediction vectors and global statistically popular motion vectors. This bias term incorporates the distance to popular vectors into the rate distortion function, allowing blocks to smoothly transition between maintaining spatial inertia and aligning with global object motions based on the specific context of each block.
3Loss of substance
If blocks inside large moving objects use locally optimal motion vectors based on neighboring blocks, then bit rate is reduced through skip coding, but the motion vectors may be far from the true motion of the object
Solution Approach 1:
The patent introduces a bias term that acts as an intermediary to guide motion vector selection toward statistically popular vectors. This bias term is incorporated into the rate distortion function and penalizes deviations from popular motion vectors, ensuring that blocks inside large moving objects select motion vectors that are both efficient for compression and accurate in representing true object motion.
Solution Approach 2:
The patent implements feedback by using statistically popular motion vectors computed from the entire frame to influence the selection of motion vectors for individual blocks. This feedback mechanism ensures that local block decisions are informed by global motion patterns, improving both accuracy and compression efficiency simultaneously.
Data Source
AI summary
A mechanism for performing a motion vector search at a current block within a current frame of video data. For each test vector position v of a plurality of test vector positions, compute a rate distortion value J(v) based on a combination of a distortion value D(v) and a bit cost value R(v,p,C). The distortion value D(v) represents a measure of distortion of the current block relative to a corresponding block Bv in the reference frame. The bit cost value R(v,p,C) depends on the test vector position v, a prediction vector p for the current block and a set C of one or more vectors c1, c2, . . . , cn. The one or more vectors c1, c2, . . . , cn are estimates (or predictions) of statistically-popular motion vectors for the current frame. The test vector position v* that gives a minimum rate-distortion value J(v*) is selected as the motion vector for the current block.