Embedded Video Denoising in Motion Compensation for Lower Bit Rate
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video coding systems face challenges in effectively removing noise from video sequences, which degrades subjective quality and reduces coding efficiency due to loss of temporal redundancy and increased bit rate requirements for noisy DCT coefficients.
Innovation Solution
The Multi-Hypothesis Motion Compensated Filter (MHMCF) and Embedded Optimal Denoising Filter (EODF) methods, which utilize temporal linear minimum mean square error estimation and can be seamlessly embedded into the motion compensation process, effectively reduce noise by using fewer input pixels and allowing flexible reference frame selection, including past and future frames.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If traditional purely pre-processing denoising filters are used, then noise removal is achieved, but extra computation is required on top of the video encoder
Solution Approach 1:
The patent merges the denoising filter with the motion compensation process by embedding the filter coefficients directly into the motion compensation equations. The multi-hypothesis motion compensation framework combines multiple reference frame predictions with the denoising operation, allowing noise removal to be performed as an integrated part of the motion compensation process rather than as a separate pre-processing step, thereby eliminating extra computation.
Solution Approach 2:
The patent creates a universal motion compensation framework that simultaneously performs motion estimation, motion compensation, and denoising functions. The multi-hypothesis approach uses multiple reference frames to provide both motion prediction and noise reduction, making the system multi-functional and eliminating the need for separate denoising processing.
2Reliability
If noise is present in video sequences, then temporal redundancy is reduced, but this decreases motion estimation accuracy and increases bit rate
Solution Approach 1:
The patent applies preliminary denoising action by embedding the denoising filter within the motion compensation process itself. The filter operates on the residue frames before they are used for motion estimation and compensation, removing noise in advance and preserving temporal redundancy for more accurate motion vector calculation, thereby improving motion estimation accuracy without requiring separate pre-processing.
Solution Approach 2:
The patent introduces an intermediary denoising filter that operates between the residue frame generation and motion estimation processes. This intermediary component removes noise from the residue frames, preserving the temporal redundancy information needed for accurate motion estimation while preventing noise from degrading the motion vector accuracy.
3Measurement precision
If more pixels are used as input for denoising filters, then denoising performance improves, but computational requirements increase
Solution Approach 1:
The patent changes the parameter of input pixel requirements by designing a denoising filter that achieves optimal performance with fewer input pixels. The multi-hypothesis motion compensation framework uses temporal information from multiple reference frames to provide sufficient denoising capability with reduced spatial sampling, thereby lowering computational energy requirements while maintaining or improving denoising performance.
Data Source
AI summary
An optimal denoising method for video coding. This method makes use of very few pixels and linear operations, and can be embedded into the motion compensation process of video encoders. This method is simple and flexible, yet offers high performance and produces appealing pictures.