White Balancing with Illumination Decomposition for Mixed Lighting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional white balance methods fail to accurately account for multiple illumination sources in an image capture environment, leading to unsatisfactory white balancing results.
Innovation Solution
A processor-implemented method using a neural network to generate a feature map, iteratively update slot vectors, and calculate mixture maps and illumination color vectors to create an illumination map for improved white balancing in images with multiple illumination sources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional white balance methods are used, then the processing is simple and fast, but the white balancing accuracy deteriorates when multiple illumination sources are present
Solution Approach 1:
The patent segments the illumination in the image into multiple independent illumination sources, each represented by a slot vector. The white balance processing is divided into separate branches, one for each illumination source, allowing independent estimation and correction of each source's chromaticity. This segmentation enables accurate handling of multiple illuminations while maintaining manageable computational complexity through structured processing.
Solution Approach 2:
The patent introduces a new dimensional framework by representing illuminations as vectors in a chromaticity space and using attention mechanisms that operate across multiple dimensions (spatial, channel, and illumination source dimensions). This dimensional expansion allows the system to capture complex illumination relationships that conventional single-dimension methods cannot handle, improving accuracy without proportionally increasing complexity.
2Adaptability or versatility
If conventional white balance assumes single illumination, then the computational load is low, but the adaptability to real-world environments deteriorates
Solution Approach 1:
The patent creates a universal white balance framework that can handle any number of illumination sources through the slot attention mechanism. The same architectural components (encoders, attention mechanisms, slot vectors) serve multiple functions: extracting features, representing different illumination sources, and computing corrections. This multi-functionality provides high adaptability to various illumination scenarios without requiring separate specialized processing for each case, optimizing computational energy usage.
3Measurement precision
If multiple illumination sources are accounted for, then white balancing accuracy improves, but the device complexity increases
Solution Approach 1:
The patent applies local quality by allowing different parts of the image to be associated with different illumination sources through spatial attention maps. Each slot vector and its corresponding attention map capture local illumination characteristics specific to certain image regions. This localized processing improves chromaticity estimation accuracy for each illumination source while keeping the overall model complexity manageable through region-specific rather than globally uniform processing.
Data Source
AI summary
An electronic device includes a processor configured to perform operations including inputting an image into an encoder to generate a feature map including information about an illumination present in the input image, iteratively updating a plurality of slot vectors using the calculated feature map to calculate a plurality of predicted illumination vectors, calculating, using the calculated plurality of predicted illumination vectors, a plurality of mixture maps representing respective effects of a plurality of virtual illuminations on pixels in the input image and a plurality of illumination color vectors representing respective color values of the plurality of virtual illuminations, and generating an illumination map using the calculated plurality of mixture maps and the calculated plurality of illumination color vectors.