Neural Network Lighting Model Application for Image Quality
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Images captured by computing devices often have imperfect lighting, which can be inconsistent, too bright or too dark, and may include undesirable tint, making it difficult to adjust the lighting of already-obtained images effectively.
Innovation Solution
A convolutional neural network is trained using confidence learning to apply a lighting model to input images, allowing for the adjustment of lighting conditions such as direction, intensity, and color, enabling the generation of output images with improved lighting based on user preferences or desired lighting models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional image correction methods are used, then simple artifacts like red-eye can be removed, but complex lighting issues such as inconsistent lighting, brightness, and tint cannot be effectively adjusted
Solution Approach 1:
The patent replaces traditional mechanical/image-processing correction methods with a neural network-based system. The neural network learns complex lighting patterns and transformations from training data, enabling it to effectively adjust various lighting conditions (inconsistent lighting, brightness, tint) that traditional methods cannot handle, while maintaining reliable and natural-looking results.
2Adaptability or versatility
If a neural network is trained to apply lighting models, then complex lighting adjustments can be made, but training complexity and computational resources increase
Solution Approach 1:
The patent applies preliminary action by pre-training the neural network on extensive datasets containing various lighting conditions and transformations before deployment. This offline training phase allows the network to learn complex lighting patterns and transformations in advance, so that during actual use, the network can quickly and efficiently apply appropriate lighting adjustments without requiring complex real-time computations.
Data Source
AI summary
Apparatus and methods related to applying lighting models to images of objects are provided. A neural network can be trained to apply a lighting model to an input image. The training of the neural network can utilize confidence learning that is based on light predictions and prediction confidence values associated with lighting of the input image. A computing device can receive an input image of an object and data about a particular lighting model to be applied to the input image. The computing device can determine an output image of the object by using the trained neural network to apply the particular lighting model to the input image of the object.