Auto White Balancing Using Eigen-Illuminant Projection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional auto white balancing methods in image signal processors are calculation intensive and generally less accurate, failing to effectively remove color biases in captured images.
Innovation Solution
The implementation of an image sensor system that includes a memory for storing eigen-illuminant images, an image sensing module, and an image signal processor. The processor generates chrominance channels, performs homogeneous region segmentation, and projects regions of interest onto eigen-illuminant images to determine gray color pixels, using machine learning algorithms to generate these images from a training set.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional auto white balancing methods are used, then the implementation is simpler, but the calculation complexity is high and accuracy is low
Solution Approach 1:
The patent pre-generates eigen-illuminant images through machine learning training on a dataset of images captured under various lighting conditions. These eigen-illuminant images are stored in memory and serve as reference templates for white balancing. During actual image processing, the system performs homogeneous region segmentation to identify gray color pixels, then compares these pixels against the pre-computed eigen-illuminant images to determine white balance gains. This preliminary computation approach shifts the heavy calculation burden from real-time processing to an offline training phase, thereby reducing real-time computational complexity while maintaining high accuracy.
2Reliability
If traditional auto white balancing methods are used, then the processing speed is faster, but the color bias removal is insufficient
Solution Approach 1:
The patent replaces traditional iterative mathematical optimization methods with a machine learning-based approach. During the training phase, the system learns optimal white balance characteristics by processing a large dataset of images under various illuminants. The learned eigen-illuminant images and their corresponding white balance gains are stored for rapid lookup during actual processing. This substitution of heavy mechanical computation with pre-learned models enables both high reliability in color bias removal and improved processing efficiency.
3Measurement precision
If machine learning algorithms are used to generate eigen-illuminant images, then the white balancing accuracy is improved, but the memory requirement increases
Solution Approach 1:
The patent extracts only the essential white balance information from the training dataset by computing and storing eigen-illuminant images and their corresponding white balance gains. Rather than storing entire training images or complex model parameters, the system extracts and retains only the critical eigen-illuminant representations needed for white balancing. This extraction approach significantly reduces memory requirements while preserving the accuracy benefits of machine learning.
Data Source
AI summary
This application describes systems and methods for detecting depth in deep trench isolation with semiconductor devices using test key transistors. An method comprises: capturing, by an image sensor, an image; generating a plurality of chrominance channels by converting the image into luminance-chrominance space; performing homogeneous region segmentation on the plurality of chrominance channels to generate one or more regions of interest in the plurality of chrominance channels; and projecting the regions of interest onto eigen-illuminant images to determine gray color pixels on the image, wherein the eigen-illuminant images are generated via performing a machine learning algorithm on a training set of images captured by the image sensor.