Motion Compensated Noise Reduction Using Adaptive Patch Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional motion compensated noise reduction (MCNR) algorithms are heavily dependent on accurate motion compensation results and lack a mechanism to select perfect matching patches, leading to suboptimal noise reduction performance.
Innovation Solution
A method and apparatus for MCNR that perform motion estimation, block matching, motion detection, and noise reduction operations using a threshold curve to generate a target patch with reduced noise, incorporating a similarity calculation and patch searching to determine the number of second patches based on motion values, and filtering noise in the frequency domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional motion compensated noise reduction (MCNR) is used with high dependence on motion compensation accuracy, then motion compensation can be performed, but noise reduction performance is suboptimal due to inability to select perfect matching patches
Solution Approach 1:
The patent applies dynamics by making the patch selection process adaptive rather than static. The system dynamically adjusts which patches to use for noise reduction by calculating motion values and comparing them against threshold curves, allowing the algorithm to adaptively select the most appropriate patches based on actual motion characteristics in different regions of the image.
Solution Approach 2:
The patent changes parameters by introducing motion value calculations and threshold curve comparisons. Instead of relying solely on motion compensation results, the system calculates motion values for different patches and uses threshold curves to determine which patches meet the criteria for perfect matching, thereby improving noise reduction performance through parameter-based selection.
2Measurement precision
If block matching algorithm is used to compare macroblocks, then motion estimation can be obtained, but the algorithm cannot determine whether selected patches are perfect matching
Solution Approach 1:
The patent implements feedback by calculating motion values for each patch and comparing them against threshold curves. This feedback mechanism allows the algorithm to evaluate the quality of motion compensation results and determine whether patches represent perfect matching, thereby improving patch matching accuracy through a systematic evaluation process.
Solution Approach 2:
The patent substitutes the conventional block matching mechanism with a motion value-based selection mechanism. Instead of relying solely on traditional block matching algorithms, the system uses calculated motion values and threshold curve comparisons to determine patch matching quality, replacing the mechanical block matching process with a more sophisticated evaluation approach.
3Reliability
If conventional MCNR highly depends on accurate motion compensation results, then motion transformation can be synthesized, but noise reduction performance deteriorates when motion compensation is inaccurate
Solution Approach 1:
The patent applies beforehand cushioning by introducing a threshold curve evaluation mechanism before final noise reduction is applied. The system calculates motion values and compares them against threshold curves to identify patches that meet the criteria for perfect matching, providing a cushioning layer of verification that protects against using inaccurate motion compensation results in the noise reduction process.
Data Source
AI summary
An apparatus for motion compensated noise reduction for input images is provided. The motion estimation and motion compensation circuit performs a motion estimation operation and a motion compensation operation on a current image and a previous image to obtain a first patch. The block matching operation circuit performs a block matching operation on the current image and the previous image to obtain a second patch. The motion detection circuit performs a motion detection operation on a target patch according to the first patch and the second patch to output a set of third patches. The current image includes the target patch. The noise reduction circuit performs a noise reduction operation on the set of third patches according to a threshold curve, so as to generate the target patch that the noise is reduced. A method for motion compensated noise reduction for input images is also provided.