Mixed-Illuminant Digital Image White Balancing with Neural Weighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing techniques struggle to accurately white balance digital images with mixed illuminants, often resulting in undesirable color tints due to correcting for only a single illumination source, which is not reflective of the actual lighting conditions in real scenes.
Innovation Solution
A method and system that utilizes a deep neural network to learn local weighting maps by processing a digital image with multiple predefined white balance settings, applying polynomial kernel functions, and blending these images to achieve accurate white balance, without requiring illuminant estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If traditional single-illuminant white balance correction is applied, then processing complexity is reduced, but color accuracy deteriorates in mixed lighting conditions
Solution Approach 1:
The image is segmented into multiple regions with different illuminants using spatial segmentation and color space transformation. The algorithm divides the image into foreground and background regions, then applies separate white balance corrections to each region based on their respective illuminant characteristics, enabling accurate color correction in mixed lighting without excessive complexity
Solution Approach 2:
Different white balance settings are applied to different spatial regions of the image based on local illuminant characteristics. The system determines dominant illuminants for different regions and applies region-specific color temperature adjustments, ensuring that each area is corrected according to its actual lighting conditions rather than applying a uniform correction across the entire image
2Measurement precision
If multiple white balance settings are processed and blended, then color accuracy improves in mixed illuminants, but processing time increases
Solution Approach 1:
The system pre-calculates multiple white balance settings (different color temperatures) and prepares them in advance. During image processing, these pre-computed settings are quickly blended based on the detected illuminant proportions, avoiding the need for real-time iterative optimization and significantly reducing processing time while maintaining color accuracy
Solution Approach 2:
The algorithm changes the color temperature parameter across multiple predefined white balance settings and blends them using learned weights. By varying this single parameter systematically and using efficient blending operations, the system achieves accurate color correction without requiring complex multi-parameter optimization, thus reducing processing time
3Measurement precision
If deep neural network is used to learn weighting maps, then white balance accuracy improves, but device complexity increases
Solution Approach 1:
A deep neural network is introduced as an intermediary component that learns to predict optimal white balance weights from image features. The network takes downsampled image patches as input and outputs weighting coefficients for blending multiple white balance settings, achieving high accuracy while keeping the overall system architecture modular and manageable through this intermediate learning layer
Data Source
AI summary
A system and method for white balancing a digital image. The method including downsampling the digital image to generate a downsampled image; processing the downsampled image with a plurality of preset white balance settings to generate a plurality of white balanced downsampled images; processing the input image at a fixed white balanced setting to produce an initial image; inputting the white balanced downsampled images to a deep neural network to generate a weighting map, the weighting map including weights of the preset white balance settings at windows of the downsampled images; generating a white balanced output image by applying the weighting map to the initial image; and outputting the white balanced output image.