Image Processing Apparatus Abnormal Pixel Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing methods using the Non-Local Bayes algorithm can result in abnormal pixel values during noise reduction, leading to failures in output images due to the calculation of inverse covariance matrices.
Innovation Solution
An image processing apparatus and method that includes setting multiple patches, calculating a target patch and covariance matrix, performing first and second corrections on pixels using the covariance matrix, determining abnormal pixels, and generating an output image based on corrected pixels to avoid failures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If noise reduction processing is performed using the Non-Local Bayes algorithm with covariance matrix calculation, then noise reduction effectiveness is improved, but abnormal pixel values may occur causing output image failure
Solution Approach 1:
The patent applies preliminary action by performing abnormality determination on pixels before they are used to generate the final output image. The determination unit checks each corrected pixel for abnormalities using statistical criteria (such as whether the pixel value falls within a reasonable range based on the covariance matrix properties) before the generation unit creates the final image. This preventive check ensures that abnormal values are identified and handled before they can cause output image failure.
Solution Approach 2:
The patent introduces an intermediary mechanism between the first correction unit and the generation unit. The determination unit acts as an intermediary that mediates the pixel values by identifying abnormalities and triggering the second correction unit when needed. This intermediary layer ensures that only valid pixel values are passed to the generation unit, preventing output image failure while maintaining the noise reduction benefits of the covariance matrix approach.
2Productivity
If a single correction process is used to reduce noise, then processing speed is improved, but abnormal pixel values may remain causing image failure
Solution Approach 1:
The patent applies segmentation by dividing the correction process into two distinct stages: the first correction unit performs initial noise reduction using the covariance matrix, and the second correction unit performs corrective processing on identified abnormal pixels. This segmentation allows the system to maintain high processing speed through the efficient first correction while ensuring reliability through the targeted second correction only when abnormalities are detected, rather than applying a single slow correction to all pixels.
Solution Approach 2:
The patent implements partial action by applying the second correction process only to the specific pixels that are determined to be abnormal, rather than correcting all pixels. The determination unit identifies only the problematic pixels using statistical criteria, and the second correction unit processes only those specific pixels. This partial correction approach maintains processing efficiency while ensuring that abnormal values are corrected, preventing output image failure without the overhead of correcting every pixel.
Data Source
AI summary
There is provided an apparatus including a first correction unit configured to acquire a first corrected pixel using the covariance matrix, a determination unit configured to determine whether the first corrected pixel is abnormal, a second correction unit configured to acquire a second corrected pixel by performing a second correction on a pixel at a position of the first corrected pixel determined to be abnormal.