Image Processing Boundary Identification for Motion Regions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Temporal noise reduction methods struggle to accurately determine the boundary between motion and non-motion regions in image frames, leading to artifacts due to unclear pixel differences.
Innovation Solution
An image processing method that compares current and previous image frames to obtain dynamic parameter values and boundary retention values, using these to refine the identification of motion and non-motion regions by adjusting luma or chroma values at the boundary, thereby enhancing accuracy and removing artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If temporal noise reduction is used to compare current and previous image frames, then noise in non-motion regions is reduced, but the boundary between motion and non-motion regions cannot be accurately determined, resulting in artifacts
Solution Approach 1:
The patent segments the image frame into motion regions and non-motion regions by comparing pixel differences between current and previous frames. By dividing the image processing into region-specific operations, it applies different noise reduction strategies to different areas, thereby accurately identifying boundaries while reducing noise in appropriate regions.
Solution Approach 2:
The patent applies different processing qualities to different regions: in motion regions, it preserves detail and avoids noise reduction that would blur moving objects, while in non-motion regions, it applies stronger noise reduction. This local differentiation resolves the contradiction by making boundary identification more accurate and reducing artifacts.
2Productivity
If a mask or sampling window is used to determine motion regions by comparing pixel differences, then motion regions can be identified, but when pixel difference is between preset value and zero, the boundary cannot be identified, creating artifacts
Solution Approach 1:
The patent changes the parameter threshold for boundary detection by introducing a dynamic threshold mechanism. Instead of using a fixed preset value, it adjusts the threshold based on local image characteristics and pixel difference distributions, enabling accurate boundary identification even when pixel differences fall between traditional threshold values.
Solution Approach 2:
The patent introduces dynamic region classification that adapts to local image content. The motion/non-motion region classification is not static but dynamically adjusted based on pixel difference analysis, allowing the system to accurately identify boundaries where pixel differences are intermediate values rather than applying a rigid threshold.
3Reliability
If spatial noise reduction is applied to smooth noise in the image frame, then noise is reduced, but boundaries and information in the image frame become blurred
Solution Approach 1:
The patent segments the image into motion and non-motion regions before applying noise reduction. In motion regions where boundaries are critical, it avoids or reduces spatial filtering to prevent boundary blurring, while in non-motion regions, it applies spatial noise reduction more aggressively. This segmentation-based approach resolves the contradiction between noise reduction and boundary preservation.
Solution Approach 2:
The patent applies different noise reduction intensities to different regions of the image. In regions with important boundary information or motion content, it uses lighter filtering to preserve detail, while in uniform non-motion regions, it applies stronger filtering for better noise reduction. This local quality differentiation maintains boundary clarity while reducing noise.
Data Source
AI summary
An image processing method includes: receiving a currently-input image frame and a previously-output image frame; comparing multiple first pixels corresponding to coordinates of the currently-input image frame with multiple second pixels corresponding to coordinates of the previously-output image frame, and obtaining multiple corresponding differences; obtaining multiple dynamic parameter values based on the differences and a dynamic parameter table; obtaining multiple boundary retention values based on the dynamic parameter values and a boundary operator; and obtaining multiple currently-output pixels based on the first pixels, the second pixels, and the boundary retention values. An image processing apparatus performs the image processing method, to increase accuracy of identifying a boundary adjoining a motion region and a non-motion region, and to remove an artifact of the boundary.