Neural Network Image Chromaticity Estimation Under AC Light
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Statistical and physical color constancy technologies face performance degradation when the statistical model is not appropriately matched or when the specular region in an image is insufficient, limiting their effectiveness in correcting illumination.
Innovation Solution
An image processing method that estimates specular and diffuse chromaticity using a neural network-based approach, generating a chromaticity dictionary matrix and coefficient matrix to perform color balancing and highlight removal in images captured under alternating current (AC) light environments, thereby improving color quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If statistical color constancy technology is used to estimate illumination, then the algorithm operating speed is high and complexity is low, but the performance is degraded greatly when the statistical model is not appropriately matched
Solution Approach 1:
The patent employs a neural network that dynamically adapts to different illumination conditions by learning from multiple frames captured over time. The network automatically adjusts its estimation based on the content and characteristics of each frame, resolving the contradiction between fast operation and reliable performance across varying lighting environments.
Solution Approach 2:
The system changes the estimation parameters by using a neural network that processes multiple temporal frames instead of relying on a fixed statistical model. This allows the system to maintain high speed while adapting to different illumination conditions, thereby improving reliability without sacrificing operating speed.
2Measurement precision
If physical color constancy technology is used to estimate illumination, then the estimation accuracy is improved, but the performance is degraded when the specular region in the image is insufficient
Solution Approach 1:
The neural network performs preliminary learning from multiple frames captured over time before processing the current frame. This preliminary action allows the system to build a comprehensive understanding of the scene and illumination conditions, enabling accurate estimation even when specular regions are insufficient in any single frame.
Solution Approach 2:
The neural network serves multiple functions: it processes specular regions, diffuse regions, and temporal variations simultaneously. This multi-functionality allows the system to maintain high estimation accuracy across diverse imaging conditions, including scenes with insufficient specular reflections, thereby improving adaptability.
3Reliability
If a neural network-based approach is used to estimate chromaticity, then the color quality performance is improved, but the device complexity increases
Solution Approach 1:
The patent segments the chromaticity estimation process into distinct components: processing specular chromaticity and diffuse chromaticity separately, and handling them through different pathways in the neural network. This segmentation allows for more manageable complexity while maintaining high color quality performance through specialized processing of each component.
Data Source
AI summary
An image processing method and apparatus is disclosed, where the image processing method includes receiving an image including frames captured over a time period in an illumination environment including alternating current (AC) light, estimating a specular chromaticity and a diffuse chromaticity of the image based on the frames, determining a weight of each of the specular chromaticity and the diffuse chromaticity based on a frame of the frames, and correcting the image based on the specular chromaticity, the diffuse chromaticity, the weight of the specular chromaticity, and the weight of the diffuse chromaticity.