Burst Image Restoration with Reference-Guided Motion Embedding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image restoration methods struggle to effectively enhance the quality of burst images captured under degraded conditions, such as low illuminance, due to variations caused by camera and object motion, without efficient utilization of motion information.
Innovation Solution
A processor-implemented method and apparatus that generates feature representations for individual images in a burst set, determines a reference feature representation, and uses motion-embedding feature representations to fuse and decode images, leveraging neural networks for encoding and decoding processes to improve image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If existing image restoration methods are used, then image quality can be improved, but motion information is not efficiently utilized leading to degradation under camera and object motion
Solution Approach 1:
The patent segments the image restoration process into distinct modules: motion estimation module that separates motion components, feature extraction module that processes static content, and fusion module that combines motion-compensated features. This segmentation allows each module to specialize in handling specific aspects of motion degradation, improving overall reliability under motion conditions.
Solution Approach 2:
The patent implements dynamic adaptation by adjusting the fusion weights between different burst images based on estimated motion parameters. The system dynamically selects and weights features from individual burst images according to their motion characteristics, allowing the restoration process to adapt to varying motion conditions rather than using fixed processing parameters.
2Manufacturing precision
If multiple burst images are processed to restore image quality, then more motion information becomes available, but computational complexity increases
Solution Approach 1:
The patent extracts only the essential motion information from burst images through motion estimation, separating motion parameters from detailed image content. By extracting and utilizing only the critical motion components rather than processing complete high-resolution images through all stages, the system reduces computational complexity while maintaining the ability to compensate for motion effects.
Solution Approach 2:
The patent applies partial processing by selectively processing only those burst images and feature regions that contain significant motion information or contribute most to restoration quality. The fusion module selectively combines features from individual bursts based on motion characteristics, avoiding full processing of all burst images at all processing stages, thus reducing overall computational burden.
Data Source
AI summary
A method and apparatus for image restoration based on burst images. The method includes generating a plurality of feature representations corresponding to individual images of a burst image set by encoding the individual images, determining a reference feature representation from among the plurality of feature representations, determining a first comparison pair including the reference feature representation and a first feature representation of the plurality of feature representations, generating a first motion-embedding feature representation of the first comparison pair based on a similarity score map of the reference feature representation and the first feature representation, generating a fusion result by fusing a plurality of motion-embedding feature representations including the first motion-embedding feature representation, and generating at least one restored image by decoding the fusion result.