Adaptive Video Denoiser with Moving Object Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video noise reduction filters (VNRs) are processor-intensive and not suitable for real-time video conferencing applications on mobile devices, as they often rely on codec-specific internal mechanisms, introducing delays and visual artifacts, and are not codec-independent.
Innovation Solution
A codec-independent adaptive noise filtering method that uses block-based temporal denoising and moving object detection, where source frames are preprocessed to identify static and moving blocks, applying noise filtering only to static blocks, and adjusting filtering thresholds based on noise level and block position, enabling real-time noise reduction without relying on motion vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional video noise reduction filters are used to reduce noise in captured video, then video quality is improved, but processing time increases and real-time performance deteriorates
Solution Approach 1:
The patent divides the video frame into multiple blocks and processes each block independently using block-based temporal filtering. This segmentation allows parallel processing of different blocks, reducing overall processing time while maintaining effective noise reduction in each block region.
Solution Approach 2:
The patent applies noise filtering selectively based on motion detection results. Instead of filtering the entire frame uniformly, it applies filtering only to static blocks where noise reduction is beneficial, while skipping moving blocks to avoid artifacts and reduce processing load.
2Object-affected harmful factors
If codec-specific internal noise filters are used, then noise reduction effectiveness is improved, but adaptability to different codecs deteriorates
Solution Approach 1:
The patent implements a universal external noise filter that operates independently of any specific video codec. It processes video data in a codec-agnostic manner, making it compatible with multiple codecs including VP8, VP9, and H.264, thus achieving multi-functionality across different coding standards.
Solution Approach 2:
The patent introduces an external noise filter as an intermediary component between the video capture device and the codec encoder. This mediator processes the raw video data before encoding, eliminating dependency on codec-specific internal filters and enabling codec independence.
3Measurement precision
If motion vectors from codec are used for noise filtering, then filtering accuracy is improved, but processing complexity and delay increase
Solution Approach 1:
The patent extracts only the essential motion information needed for noise filtering by performing simple block-based motion estimation using sum of absolute differences (SAD). It takes out only the necessary motion data rather than relying on the complete motion vector set from the codec, reducing processing complexity while maintaining sufficient accuracy.
Solution Approach 2:
The patent uses a simplified, computationally inexpensive motion estimation method that requires minimal processing resources. This lightweight approach provides sufficient motion detection accuracy for noise filtering purposes without the high computational cost of more sophisticated motion estimation algorithms.
4Object-affected harmful factors
If strong noise filtering is applied to all blocks, then noise reduction is improved, but visual artifacts in moving regions increase
Solution Approach 1:
The patent applies different filtering strategies to different regions of the video frame based on local motion characteristics. Static blocks receive strong noise filtering, while moving blocks are either skipped or receive reduced filtering, ensuring that each region is processed according to its specific quality requirements and avoiding artifacts in moving areas.
Data Source
AI summary
A method and apparatus for adaptive denoising of source video in a video conference application is provided. Source video frames received from a video capture device are divided into a plurality of 16×16 blocks. For each source block, a moving object detection process and a noise estimation process are performed. Then temporal denoising is adaptively applied to the blocks of the source frame based on the noise estimation and moving object detection. The adaptively filtered blocks are provided to an output frame and forwarded to a coding module for encoding.