Temporal Noise Reduction Filter With Adaptive Frame Blending
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing temporal noise reduction techniques in low light video capture often result in noisy images due to overreliance on previous frames with low variance, leading to undesirable noise and slow convergence.
Innovation Solution
A filter incorporating a Kalman gain filter with Sigmoid interpolation is used to account for noise variance and temporal differences between frames, determining optimal blending ratios to reduce noise effectively and quickly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a high blending ratio is used when there is low temporal difference between frames, then temporal noise reduction is improved, but noise is introduced due to overreliance on previous frame
Solution Approach 1:
The blending ratio is made dynamic and adaptive rather than fixed. The system automatically adjusts the blending ratio based on the measured temporal difference between frames, using high blending ratios when temporal difference is low and low blending ratios when temporal difference is high, thereby optimizing noise reduction while avoiding overreliance on previous frames
Solution Approach 2:
The system measures the temporal difference between current and previous frames as feedback, then uses this measurement to adjust the blending ratio accordingly. This closed-loop approach ensures that the blending ratio adapts to actual temporal variations, preventing both excessive smoothing and insufficient noise reduction
2Object-generated harmful factors
If a low blending ratio is used when there is high temporal difference between frames, then noise from previous frame is reduced, but temporal noise reduction effectiveness decreases
Solution Approach 1:
The blending ratio dynamically adapts to the temporal difference magnitude. When temporal difference is high, the system automatically reduces the blending ratio to prevent propagating noise from previous frames, while when temporal difference is low, it increases the blending ratio to maximize noise reduction effectiveness
Solution Approach 2:
The system uses temporal difference measurement as feedback to control the blending ratio. This feedback mechanism ensures that the blending ratio is optimally adjusted based on actual frame-to-frame variations, preventing noise propagation when temporal difference is high while maintaining effective noise reduction when temporal difference is low
3Object-generated harmful factors
If traditional temporal noise reduction is used, then noise is reduced, but convergence is slow
Solution Approach 1:
The system performs preliminary measurement of temporal difference between frames before applying the blending operation. This preliminary action allows the system to pre-determine the optimal blending ratio, enabling faster convergence by avoiding iterative adjustment and directly applying the appropriate blending weight from the start
Solution Approach 2:
The system changes the blending ratio parameter based on measured temporal difference. By adjusting this critical parameter dynamically rather than using a fixed value, the system achieves faster convergence to optimal noise reduction results, reducing the time required to reach effective noise suppression
Data Source
Figure 1
Figure 2A~2B
Figure 3
AI summary
Methods, systems, and apparatus, including computer programs encoded on a computer storage medium, for using a filter for temporal noise reduction. In some implementations, image data for a series of frames, including a first input frame followed by a second input frame, of a video is obtained. A first output frame resulting from noise reduction processing for the first input frame and a measure of variance associated with a portion of the first output frame is obtained. An interpolation setting for noise reduction processing of a portion of the second input frame is determined. A second output frame is generated by interpolating the portion of the second input frame with the corresponding portion of the first output frame.