Pixel-Wise Noise Correction Using Block Representative Pixels
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional noise correction methods in imaging devices correct noise on a row-wise or column-wise basis, failing to account for variations within a single row and often rely on defective pixels for noise sampling, leading to aberrations like row-banding and inaccurate noise detection.
Innovation Solution
Implementing a pixel-wise noise correction method that uses adjacent or nearby pixels as reference pixels to determine noise levels, allowing for more accurate noise correction by comparing image and reset values between active and reference pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If row-wise or column-wise noise correction is used, then the correction process is simple, but it fails to account for variations within a single row and relies on defective pixels leading to row-banding artifacts
Solution Approach 1:
The patent segments the pixel array into multiple blocks, where each block contains a subset of pixels. Noise correction is performed independently for each block using representative pixels from that specific block, rather than applying a uniform correction across entire rows or columns. This segmentation enables pixel-wise noise correction without requiring complex global processing, as each block can be processed relatively independently.
Solution Approach 2:
The patent applies local quality by selecting representative pixels from specific blocks within the pixel array to characterize noise for that local region. Each block has its own representative pixels and correction values, allowing noise correction to adapt to local variations in the pixel array. This local approach captures within-row variations that global row-wise correction misses, while avoiding the complexity of processing every single pixel globally.
2Ease of manufacture
If optical black pixels are used for noise sampling, then the method is simple to implement, but it provides inaccurate noise detection and causes row-banding artifacts
Solution Approach 1:
The patent extracts representative pixels from actual image-bearing blocks rather than relying on separate optical black pixels. By taking out pixels that are actually part of the image data and using them as representatives for their respective blocks, the method achieves accurate noise characterization without the inaccuracies associated with optical black pixels, which do not experience the same signal paths and noise characteristics as active pixels.
Solution Approach 2:
The patent uses representative pixels that copy the actual noise characteristics of their block by being selected from within the same block. Instead of copying noise values from distant optical black pixels, the representative pixel is chosen from the same block and thus naturally copies the local noise properties including row-specific variations, providing more accurate noise detection.
3Measurement precision
If representative pixels are selected from blocks, then accurate noise correction is achieved, but the selection process becomes more complex
Solution Approach 1:
The patent divides the pixel array into multiple blocks and selects representative pixels from each block independently. This segmentation simplifies the selection process by limiting the search space to smaller blocks rather than the entire array, making the selection manageable while still achieving accurate local noise characterization for each block.
Solution Approach 2:
The patent selects only a subset of pixels (representative pixels) from each block rather than processing all pixels. This partial action approach achieves sufficient noise correction accuracy by using enough representative samples from each block without the excessive complexity of analyzing every single pixel, balancing accuracy and computational feasibility.
Data Source
AI summary
Methods and apparatuses providing pixel-wise noise correction using pixels to provide reference values during pixel readout operations.