Auto White Balance Algorithm Using Block Variance and Patch Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing devices face challenges in accurately and efficiently adjusting white balance under varying illuminance conditions, leading to costly and complex auto white balancing methods that are not suitable for mobile devices.
Innovation Solution
An image processing device and method that calculates block variance, selects patches in a gray zone, applies mean shift operations, and combines first and second RGB gains to achieve accurate auto white balancing, reducing complexity and power consumption.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a complex auto white balance method is used to accurately detect white component, then measurement precision is improved, but device complexity increases and power consumption rises
Solution Approach 1:
The image is divided into multiple blocks, and each block is further divided into patches. This segmentation allows the algorithm to process smaller regions independently, reducing the overall computational complexity while maintaining detection accuracy through localized analysis of white components in different image regions
Solution Approach 2:
Different processing strategies are applied to different regions based on their characteristics. The patent calculates variance for each block and selectively processes blocks that meet certain criteria, applying localized gray zone detection and mean shift operations only where needed, rather than uniformly processing the entire image
2Measurement precision
If a complex auto white balance method is used to accurately detect white component, then measurement precision is improved, but power consumption increases
Solution Approach 1:
By dividing the image into blocks and patches, the computational workload is distributed and reduced compared to processing the entire image at once. This segmentation enables efficient use of computational resources, lowering power consumption while maintaining accurate white component detection through focused local analysis
Solution Approach 2:
The patent applies mean shift operation and gray zone detection only to selected blocks that meet variance criteria, rather than processing the entire image. This partial action approach reduces unnecessary computations and power consumption while still achieving accurate white balance by focusing processing on relevant regions
3Manufacturing precision
If block variance calculation and multiple patch selections are performed, then manufacturing precision is improved, but productivity decreases
Solution Approach 1:
The image processing is segmented into hierarchical levels (image → blocks → patches), allowing parallel processing of independent regions. This segmentation structure enables the system to maintain high precision through multiple selection stages while improving processing speed by distributing computations across different regions simultaneously
Solution Approach 2:
The patent performs preliminary variance calculation on blocks before conducting detailed patch analysis. Blocks that do not meet the variance threshold are excluded from further processing, performing a preliminary filtering action that reduces the number of patches requiring intensive mean shift operations, thereby improving overall processing efficiency
Data Source
AI summary
An auto white balancing method for image data includes calculating block variances for the image data and selecting blocks having a block variance higher than a block variance value. The method further includes selecting patches included in a gray zone defined at a color space, applying a corresponding block variance as a weight to the patches of the gray zone, adjusting a first mean value at the color space, and selecting first patch candidates belong to a first distribution at a luminance domain from among patches selected with reference to the adjusted first mean value. The method further includes selecting a shrunken gray zone corresponding to an anchor point in the gray zone and selecting second patch candidates having a second distribution at a color space and a luminance domain, from among patches included in the shrunken gray zone, and calculating a final gain for the image data by combining a first RGB gain and a second RGB gain respectively extracted from the first patch candidates and the second patch candidates.