Pixel Interference Compensation Using Median Filters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing methods suffer from false color issues due to pixel interference, particularly in tiny line structures, as they either compensate only one type of pixel or capture excessive surrounding pixels, leading to color deviations and bluish or reddish tendencies in images.
Innovation Solution
The method involves selecting a first color pixel and calculating compensation values using average values from horizontally and vertically adjacent pixels, employing four standard median filters to extract edge information and obtain average values for Gb and Gr pixels, which are then used to calculate compensated pixel values, effectively addressing pixel interference in regions with uniform and smooth colors as well as on edges.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If conventional compensation methods capture a great quantity of surrounding pixels to calculate average values, then the compensation coverage is improved, but false color occurs in tiny line structures
Solution Approach 1:
The patent applies local quality by differentiating compensation strategies based on local image characteristics. It identifies whether a pixel belongs to a tiny line structure or a uniform region, and applies different compensation formulas accordingly. For tiny line structures, it uses a formula that considers fewer surrounding pixels to preserve edge sharpness, while for uniform regions, it uses a formula that averages more surrounding pixels to improve compensation coverage.
2Device complexity
If only one type of pixel (Gr or Gb) is compensated, then the compensation process is simplified, but the entire image tends to be bluish or reddish
Solution Approach 1:
The patent merges the compensation processes for both Gr and Gb pixels into a unified method. It provides specific compensation formulas for both pixel types, ensuring that both green pixel types are compensated simultaneously. This prevents the color bias that occurs when only one type is compensated, while maintaining a relatively simple overall process through the use of standardized calculation procedures.
3Measurement precision
If median value extraction is used for compensation, then edge detection capability is improved, but wrong median is extracted for tiny line structures causing false color
Solution Approach 1:
The patent changes the compensation parameter selection based on the local image structure. Instead of universally applying median value extraction, it switches between different compensation approaches: using formulas with fewer surrounding pixels for tiny line structures to avoid wrong median extraction, and using more comprehensive averaging for uniform regions to maintain edge detection capability where appropriate.
Data Source
AI summary
A method for compensating a pixel of an image is provided The method includes using four median value extraction matrices to obtain the edge information of the image, where the edge information obtained by two of the matrices is Gb value, and the edge information obtained by the other two matrices is Gr value, calculating the corresponding average values of the Gb and Gr values, multiplying a differential value between the average value of the Gb value and the average value of the Gr value with a proportion value to obtain a compensation value, and adding the compensation value to or subtracting the compensation value from the original Gb and Gr values to obtain compensated Gb and Gr values.