Temporal Filtering via Non-Adjacent Motion Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional motion detection methods fail to accurately detect slow motion and motion in noisy video sequences due to noise and scene lightness changes, leading to false positives and negatives, which affect the quality of temporal filtering.
Innovation Solution
The method involves computing a motion score between a target area in a target picture and non-adjacent reference pictures, using multiple motion detection scores to control temporal filtering, and applying adaptive filtering techniques to combine target and reference pictures based on motion detection between non-adjacent frames, thereby improving motion detection robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional motion detection uses local error measures between adjacent pictures, then the detection is computationally simple, but it produces false positives and negatives due to noise and lightness changes
Solution Approach 1:
The patent divides the motion detection process into multiple stages: first performing a coarse motion detection between non-adjacent reference pictures to identify potential motion regions, then performing refined motion detection in those specific regions. This segmentation allows the system to achieve high detection accuracy without applying complex algorithms to the entire picture, thus resolving the contradiction between reliability and complexity.
Solution Approach 2:
The patent introduces a temporal dimension by comparing non-adjacent pictures (reference picture n-k with target picture n) in addition to adjacent pictures. This multi-temporal-dimension approach provides more reliable motion detection by observing motion patterns over longer time intervals, reducing false positives from transient noise while maintaining computational feasibility through selective application.
2Reliability
If temporal filtering is applied strongly to reduce noise, then noise reduction is improved, but motion artifacts increase when motion is present
Solution Approach 1:
The patent implements dynamic temporal filtering where the filter strength adapts based on detected motion. The system first performs motion detection to identify regions with motion, then applies strong temporal filtering only to stationary regions and weak or no filtering to motion regions. This dynamic adaptation resolves the contradiction by making filter strength conditional on local motion characteristics.
Solution Approach 2:
The patent applies different filtering strengths to different spatial regions based on local motion detection results. Regions identified as stationary receive strong temporal filtering for effective noise reduction, while regions identified as having motion receive reduced or no filtering to preserve motion details and avoid artifacts. This local differentiation resolves the contradiction between noise reduction and artifact prevention.
3Reliability
If motion detection covers a wide area to improve robustness, then detection robustness is improved, but detection precision for local motion decreases
Solution Approach 1:
The patent segments the detection process into wide-area preliminary detection between non-adjacent pictures to establish robust motion patterns, followed by localized refined detection in identified regions. This two-stage segmentation allows the system to benefit from both wide-area robustness and local precision without compromising either aspect.
Solution Approach 2:
The patent adds a temporal dimension by incorporating comparisons between non-adjacent pictures (n-k to n) alongside adjacent picture comparisons. This multi-temporal-尺度 approach allows wide temporal baseline for robustness while maintaining local precision through region-specific analysis, resolving the contradiction between robustness and precision.
Data Source
AI summary
A method for temporal filtering based on motion detection between non-adjacent pictures. The method may compute a motion score by motion detection between a target area in a target picture and a first area in a non-adjacent one of a plurality of reference pictures; and temporal filter the target area with a second area in an adjacent one of the reference pictures based on the motion score to generate a filtered area in a filtered picture. At least one of (i) the motion score and (ii) the generation of the filtered area may be controlled by one or more gain settings in a circuit.