Multispectral Imaging for Deep Learning White Balance Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image sensors struggle with accurately estimating illumination conditions due to overlapping wavelength information in RGB spaces, leading to improper white balance and incorrect color representation, especially when assumptions like the Gray World Assumption are not satisfied.
Innovation Solution
Employing a multispectral image sensor capable of acquiring images in multiple channels across a wide wavelength band (10 nm to 1,000 nm) and utilizing a deep learning network to estimate illumination information, enabling precise color conversion and white balance correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If RGB image is captured and white balance is performed by analyzing information contained in the RGB image, then the process is simple and fast, but the color accuracy deteriorates when illumination conditions are complex or Gray World Assumption is not satisfied
Solution Approach 1:
The patent transitions from 3-channel RGB space to multispectral imaging with multiple wavelength bands, adding dimensional information about spectral characteristics. This enables more accurate illumination estimation by analyzing reflectance patterns across different wavelengths, resolving the contradiction between simple processing and accurate color representation under complex lighting conditions.
Solution Approach 2:
The patent introduces an illumination estimation module that acts as an intermediary between image capture and white balance processing. This module estimates illumination conditions by analyzing spectral information and generates correction factors, serving as a mediator that improves color accuracy without significantly increasing processing complexity.
2Measurement precision
If multispectral images are acquired and deep learning network is used to estimate illumination information, then color accuracy and white balance improve, but device complexity increases
Solution Approach 1:
The patent employs a deep learning network that has been pre-trained offline to perform illumination estimation. The network is trained in advance on large datasets to learn spectral patterns and illumination characteristics, so during actual operation, it can quickly estimate illumination conditions without requiring complex real-time computations, thus reducing operational device complexity while maintaining high accuracy.
Solution Approach 2:
The patent designs a unified multispectral imaging system that captures images across multiple wavelength bands and simultaneously performs both illumination estimation and reflectance analysis. This multi-functional approach consolidates multiple tasks into a single system, managing complexity by making the device perform multiple functions rather than requiring separate specialized systems.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Accurately separates illumination components from reflectance, allowing for improved color representation and white balance, enhancing image quality and stability in applications like object recognition and aesthetic processing.
Implementation Method 1
Image sensors receive light incident from an object and photoelectrically convert the received light into an electrical signal
Data Source
AI summary
An image acquisition apparatus includes: a multispectral image sensor configured to acquire images in at least four channels based on a second wavelength band of about 10 nm to about 1,000 nm; and a processor configured to estimate illumination information of the images by inputting the images of at least four channels to a deep learning network trained in advance, and convert colors of the acquired images using the estimated illumination information.