Video Noise Reduction Using Human Visual System Weighting Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video compression technologies face challenges in accurately identifying and reducing video noise and difficult-to-track visual details, which affects the quality of compressed video images and requires more bits for transmission.
Innovation Solution
A system that uses a model of the human visual system to create a pixel map and determine a weighting map from the motion-compensated difference between original and reference pictures, allowing for adaptive noise reduction and detail reduction, thereby reducing the energy of difficult-to-compress and difficult-to-perceive pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If video noise and difficult-to-track visual details are reduced to decrease compression bits, then the number of bits required for encoding is reduced, but important visual details may be removed causing perceptual degradation
Solution Approach 1:
The patent applies local quality by creating spatial and temporal weighting maps that assign different importance weights to different regions and time points in the video. The spatial weighting map identifies regions with low visual importance (e.g., uniform areas, textured regions) where noise reduction can be applied more aggressively, while the temporal weighting map identifies time points where motion is low and noise is more prominent. This localized approach allows selective noise reduction in less important areas while preserving important visual details, resolving the contradiction between bit reduction and quality maintenance.
Solution Approach 2:
The patent changes the parameter of noise reduction strength dynamically based on spatial and temporal characteristics. By computing weighting maps that reflect human visual system properties and motion information, the system adapts the noise reduction parameter (weighting factor) to different regions and time points. This parameter adaptation allows stronger noise reduction where it is imperceptible and weaker reduction where it would degrade quality, thus reducing bits while maintaining perceived video quality.
2Productivity
If aggressive noise reduction is applied to reduce video entropy, then compression efficiency improves, but accurate identification of video noise becomes difficult and important details may be lost
Solution Approach 1:
The patent applies preliminary action by computing spatial and temporal weighting maps before the actual noise reduction process. These weighting maps are derived from the original video frames and motion information, providing a guide for subsequent noise reduction operations. By establishing this preliminary spatial-temporal importance map, the system identifies which regions and time points are suitable for noise reduction without losing important details, thereby improving both noise identification accuracy and compression efficiency.
Solution Approach 2:
The patent introduces weighting maps as an intermediary between the original video and the noise reduction process. These maps serve as a mediator that encodes spatial and temporal importance information, guiding the noise reduction algorithm to apply different strengths in different regions and time points. This intermediary structure allows the system to accurately distinguish between noise and important visual details, enabling effective noise reduction while preserving quality.
Data Source
AI summary
A system includes a data storage configured to store a model human visual system, an input module configured to receive an original picture in a video sequence and to receive a reference picture, and a processor. The processor is configured to create a pixel map of the original picture using the model human visual system. A first layer is determined from the pixel map. A weighting map is determined from a motion compensated difference between the original picture and the reference picture. A processed picture is then determined from the original picture using the weighting map and the first layer.