Motion-Adaptive Image Stabilization Using Spatial Temporal Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Camera motion or object motion in low light conditions causes motion blur in images captured by solid-state cameras, and existing image stabilization techniques exacerbate noise when shortening exposure time to mitigate blur.
Innovation Solution
A motion-adaptive system employing spatial and temporal filtering of pixel signals from multiple frames captured with short exposure times, using a recursive process that requires only one additional image buffer, to produce a stabilized image while minimizing noise amplification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If exposure time is shortened to reduce motion blur, then motion blur is mitigated, but noise is amplified due to increased gain
Solution Approach 1:
The patent divides a single long exposure into multiple short exposures (segmentation of exposure time). By capturing multiple frames with short exposure times and combining them, the system achieves the total exposure duration needed while avoiding motion blur associated with any single long exposure. The segmentation allows temporal filtering to reduce noise that appears differently across frames.
Solution Approach 2:
The patent implements continuous capture of multiple frames throughout the total exposure period, maintaining continuous useful action. Rather than a single discontinuous exposure, the system continuously captures images and combines them through temporal filtering, ensuring that the full exposure time is utilized effectively while mitigating motion blur and noise through the combination process.
2Object-affected harmful factors
If multiple frames are captured and combined, then noise is reduced through temporal filtering, but device complexity increases
Solution Approach 1:
The patent employs dynamic motion-adaptive filtering where the filtering parameters and weights are adjusted based on detected motion in each frame. Rather than using fixed filtering parameters, the system dynamically adapts the temporal filtering strength and spatial filtering characteristics according to the actual motion content, optimizing noise reduction while preserving motion details and avoiding excessive processing of static scenes.
Solution Approach 2:
The patent changes filtering parameters dynamically based on motion detection results. The filter strength, temporal weighting, and spatial filtering parameters are adjusted according to the measured motion magnitude and direction in each frame sequence. This parameter adaptation allows the system to achieve effective noise reduction in static regions while preserving motion fidelity in moving regions, reducing overall processing complexity compared to uniform strong filtering.
3Illumination intensity
If gain is increased to compensate for short exposure time, then exposure is sufficient, but noise amplification occurs
Solution Approach 1:
The patent segments the total exposure into multiple short exposures, each capturing sufficient light signal without requiring excessive gain amplification. By distributing the total exposure requirement across multiple frames and combining them through temporal filtering, the system achieves the needed exposure level while keeping individual frame gains moderate, thereby reducing noise amplification compared to a single long exposure with high gain.
Solution Approach 2:
The continuous capture and combination of multiple frames maintains sufficient exposure levels throughout the capture period. The temporal filtering process accumulates signal information continuously while averaging out random noise, achieving the required illumination intensity and exposure level without relying on high gain amplification that would magnify noise in any single frame.
Data Source
AI summary
A method and apparatus for image stabilization while mitigating the amplification of image noise by using a motion adaptive system employing spatial and temporal filtering of pixel signals from multiple captured frames of a scene.