Neural Network Auto-White-Balancing Mixed Lighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current auto-white-balance methods fail to accurately describe mixed lighting conditions in real-life images and do not account for sensor variations across different cameras, leading to inaccurate color temperature estimation and unbalanced RGB channels.
Innovation Solution
A neural network-based auto-white balancing method that uses logarithmic ratios of red and green channel values, and blue and green channel values to create multiple 2D histograms from original, Gaussian-blurred, sharpened, and Laplacian-edge images, determining a white balancing gain to correct for sensor variations and mixed lighting conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If current mathematical models are used to estimate color temperature, then the method is simple to implement, but the accuracy of color temperature estimation deteriorates in mixed lighting conditions
Solution Approach 1:
The patent transforms the color temperature estimation problem from direct temperature prediction to predicting color balance parameters (Rg and Yg) in a CIE 1931 chromaticity diagram. This parameter transformation allows the model to handle mixed lighting conditions more effectively by working with chromaticity coordinates that are more robust to varying illumination conditions.
Solution Approach 2:
The patent introduces multiple processing dimensions by applying different image processing techniques (Gaussian blur, sharpening, edge detection) to create multiple versions of the input image. Each processed version provides different contextual information about the scene, and the model aggregates information across these dimensions to improve estimation accuracy in complex lighting scenarios.
2Adaptability or versatility
If traditional auto-white-balance methods are used, then the processing speed is fast, but the ability to correct sensor variations between cameras deteriorates
Solution Approach 1:
The patent incorporates a pre-processing stage that applies multiple image processing operations (Gaussian blur, sharpening, edge detection) before the main prediction step. This preliminary action prepares the data in various forms that help the model adapt to different sensor characteristics and lighting conditions, improving generalization across different camera devices.
Solution Approach 2:
The trained model is designed to be universally applicable across different camera sensors and lighting conditions. By training on diverse datasets and using a chromaticity-based approach rather than sensor-specific parameters, the model achieves multi-functionality that works across various device types without requiring device-specific calibration.
3Measurement precision
If multiple image processing operations are applied to create different 2D histograms, then the accuracy of white balance determination improves, but the computational time increases
Solution Approach 1:
The patent applies multiple image processing operations (Gaussian blur, sharpening, edge detection) to create enhanced versions of the input image for histogram analysis. While this increases computational effort, it provides richer feature information that significantly improves white balance accuracy, particularly in challenging lighting conditions where simple histograms would fail.
Solution Approach 2:
The patent creates multiple copies of the input image, each processed differently (original, blurred, sharpened, edge-detected). These copies are then used to generate separate 2D histograms that capture different aspects of the image data. The model aggregates information from all these histogram copies to make a more robust white balance determination.
Data Source
AI summary
A method of auto white balancing, including, receiving an original image, determining an RG logarithmic ratio of a set of red to green channel values of the original image, determining a BG logarithmic ratio of a set of blue to green channel values of the original image, determine an original two-dimensional histogram utilizing the RG logarithmic ratio and the BG logarithmic ratio, determine a Gaussian-blur two-dimensional histogram utilizing the RG logarithmic ratio and the BG logarithmic ratio, determining a sharpened two-dimensional histogram of a sharpened image utilizing the RG logarithmic ratio and the BG logarithmic ratio, determining a Laplacian-edge two-dimensional histogram of a Laplacian-edge image utilizing the RG logarithmic ratio and the BG logarithmic ratio and determining a white balancing gain utilizing a neural network based on the original 2D histogram, the Gaussian-blur 2D histogram, the sharpened 2D histogram and the Laplacian-edge 2D histogram.