Neural White Balance Editing for Post-Capture Color Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing digital cameras make it challenging to edit white balance settings after image capture, especially when incorrect settings result in color casts, impacting photographic quality and computer vision applications.
Innovation Solution
A deep learning framework using a DNN architecture with an encoder and multiple decoders allows for post-capture white balance editing of sRGB images, enabling correction and adjustment to various illumination settings.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If white balance correction is applied by image signal processing to normalize illumination effects, then color accuracy is improved, but the ability to edit white balance settings after capture is lost
Solution Approach 1:
The system performs preliminary white balance correction using ISP to ensure color accuracy, while simultaneously preserving the original unprocessed data in an intermediate representation format. This allows post-capture editing without sacrificing the color accuracy achieved by ISP correction.
Solution Approach 2:
The patent introduces an intermediate representation as a mediator between the original raw image and the final corrected image. This intermediate form retains the structure needed for white balance editing while incorporating the color accuracy improvements from ISP processing.
2Reliability
If white balance settings are adjusted during image capturing, then color cast issues can be prevented, but flexibility to change settings after capture is reduced
Solution Approach 1:
The system applies preliminary white balance correction during capture to prevent color cast issues, while simultaneously maintaining an editable intermediate representation that enables post-capture adjustment when needed.
Solution Approach 2:
The patent creates a dynamic system where white balance settings can be adjusted either at capture time or after capture, depending on user needs. The intermediate representation enables this flexibility without compromising the reliability of color cast prevention.
3Adaptability or versatility
If deep learning models are used for white balance editing, then editing flexibility is improved, but computational complexity increases
Solution Approach 1:
The system performs computationally intensive deep learning processing in advance to create the intermediate representation, enabling flexible post-capture editing without requiring complex computations at the time of editing.
Solution Approach 2:
The patent creates a copy of the image data in an intermediate representation format that preserves the necessary information for white balance editing. This copying approach enables flexible editing without repeatedly processing the original complex data.
Data Source
AI summary
An apparatus for white balance editing, includes a memory storing instructions, and at least one processor configured to execute the instructions to obtain an input image having an original white balance that is corrected by image signal processing, and obtain, using a first neural network, an intermediate representation of the obtained input image, the intermediate representation having the original white balance that is not corrected by the image signal processing. The at least one processor is further configured to execute the instructions to obtain, using a second neural network, a first output image having a first white balance different than the original white balance, based on the obtained intermediate representation.