Variable Search Window for Motion Vector Refinement
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video transcoding methods face challenges in accurately estimating output motion vectors, particularly during non-integer scaling, where the estimation is complex and computationally intensive due to varying influences of overlapping input macroblocks on output macroblocks, leading to potential image quality degradation.
Innovation Solution
A method for estimating output motion vectors involves generating candidate vectors, selecting reference vectors based on spatial weight and block activity, and establishing a variable search window according to the relativity between these vectors to refine the motion vector estimation, thereby optimizing the search window size and reducing computational complexity.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a fixed search window is used for motion vector refinement, then the estimation process is simple, but the accuracy of output motion vector estimation deteriorates during non-integer scaling
Solution Approach 1:
The patent applies dynamics by making the search window size variable rather than fixed. The search window size is dynamically adjusted based on the scaling factor, allowing the system to adapt to different scaling conditions. For non-integer scaling, a larger search window is used to capture the broader range of possible motion vectors, while for integer scaling, a smaller window suffices, thus resolving the contradiction between simplicity and accuracy.
Solution Approach 2:
The patent changes the parameter of search window size based on the scaling factor. By modifying this parameter according to the specific scaling condition (integer or non-integer), the system optimizes the balance between computational complexity and estimation accuracy. This parameter adaptation allows the search window to be appropriately sized for each scaling scenario.
2Measurement precision
If a large search window is used for motion vector refinement, then the accuracy of output motion vector estimation is improved, but the computational complexity increases
Solution Approach 1:
The system dynamically adjusts the search window size based on the scaling factor, using a large window only when necessary (non-integer scaling) and a smaller window for integer scaling. This dynamic adaptation reduces unnecessary computational complexity while maintaining accuracy where needed.
Solution Approach 2:
The search window size parameter is changed according to the scaling factor, optimizing the trade-off between accuracy and computational power. By adjusting this parameter, the system avoids the constant high computational cost of a always-large window while preventing the accuracy loss of a always-small window.
3Productivity
If motion vector refinement is performed with a fixed search window, then the processing speed is maintained, but the transcoder performance deteriorates during non-integer scaling
Solution Approach 1:
The system dynamically adjusts the search window size based on the scaling factor, ensuring optimal transcoder performance for different scaling conditions. This dynamic approach maintains high processing speed by avoiding unnecessarily large windows for integer scaling while improving reliability for non-integer scaling through appropriately larger windows.
Solution Approach 2:
The search window parameter is adapted according to the scaling factor, allowing the system to optimize the balance between processing speed and transcoder performance. This parameter change ensures that the system maintains productivity while improving reliability in specific scaling scenarios.
4Measurement precision
If overlapping input macroblocks are considered with varying weights, then the accuracy of motion vector estimation is improved, but the calculation complexity increases
Solution Approach 1:
The patent applies local quality by assigning different weights to different overlapping input macroblocks based on their spatial relationship with the output macroblock. This weighted approach focuses computational effort on the most relevant macroblocks, improving estimation accuracy while managing calculation complexity through selective weighting rather than uniform processing of all overlapping blocks.
Data Source
AI summary
A method of estimating output motion vector includes generating a group of candidate motion vectors, selecting two or more reference vectors, variably establishing a search window, and estimating an output motion vector by performing a motion vector refinement on the search window. The method selects the two or more reference vectors from the group of the candidate motion vectors and the search window is established according to a relativity among the reference vectors. A transcoder using the method includes an encoding parameter estimator configured to select the two or more reference vectors and to variably establish the search window according to the relativity between the reference vectors. Therefore, the search window may be efficiently established without increasing calculation time, so that the output motion vector may be more correctly estimated.