Motion Compensation for Noise Reduction in Low Light Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems fail to effectively reduce noise in images captured in low light conditions without amplifying noise, especially when global and local motion are present, leading to blurring effects in composite images.
Innovation Solution
A system that identifies portion motion vectors and global motion vectors in multiple frames, performs motion compensation to align frames with consistent global motion, and incorporates inconsistent portions from other frames to generate a noise-reduced output image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple frames are combined to reduce noise, then image quality improves, but blurring occurs due to global and local motion
Solution Approach 1:
The image is divided into multiple portions or blocks, and motion compensation is performed independently for each portion. This allows different motion parameters to be applied to different regions, enabling accurate alignment of static portions while preserving the motion characteristics of moving portions, thus avoiding blurring when combining multiple frames.
Solution Approach 2:
Different portions of the image are treated differently based on their motion characteristics. Static portions undergo strict motion compensation to eliminate blur, while moving portions are handled with adaptive motion parameters that preserve their motion characteristics. This local differentiation allows noise reduction through frame combination without uniform blurring across the entire image.
2Illumination intensity
If simple signal amplification is used to improve low light images, then brightness increases, but noise is also amplified
Solution Approach 1:
Multiple frames captured in low light conditions are combined through motion-compensated averaging. By aligning corresponding portions across frames using motion vectors, the signal components reinforce each other while random noise components tend to cancel out, effectively improving the signal-to-noise ratio without simple amplification.
Solution Approach 2:
The system continuously captures multiple frames and processes them in real-time to maintain optimal image quality. The continuous capture and processing allow for temporal averaging that reduces noise while preserving the continuous motion information, providing sustained improvement over single-frame amplification.
3Reliability
If motion compensation is applied to align frames, then noise reduction improves, but local motion details are lost
Solution Approach 1:
The motion compensation system uses dynamic motion parameters that are adapted to each portion of the image. Motion vectors are calculated and applied differently for each block or portion, allowing the system to track and preserve local motion characteristics while still enabling noise reduction through frame combination. This dynamic adaptation prevents loss of local motion details.
Solution Approach 2:
Motion parameters such as displacement vectors and motion models are changed and optimized for different portions of the image based on their specific motion characteristics. This parameter differentiation allows the system to apply appropriate motion compensation to each region, preserving local motion information while achieving overall noise reduction through the combination of multiple frames.
Data Source
AI summary
Systems and methods are provided for mitigating image noise. Portion motion vectors are identified for a plurality of portions of a first frame relating to a setting, the portion motion vectors indicating motion from the first frame to a second frame. A global motion vector is determined that indicates global motion spanning the first frame to the second frame. A determination is made that a first portion of the plurality of portions has a portion motion vector that is consistent with the global motion vector. Motion compensation is performed on the first frame and the second frame to align the second frame with first frame at the first portion that is consistent with the global motion vector to generate an output image that is based on both of the two frames. An inconsistent second portion of one of the frames is incorporated into the output image.