Compressed Image Upsampling With Artifact-Aware Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image upsampling methods, such as bicubic interpolation and super-resolution neural networks, fail to effectively reduce the visibility of compression artifacts in low-resolution images, leading to unsatisfactory image quality and increased computational demands.
Innovation Solution
A system comprising a feature extraction engine, reconstruction engine, and upsampling engine, utilizing a combination of convolutional neural networks and atrous convolutions, to identify and mitigate compression artifacts while upsampling images, thereby enhancing image quality and reducing computational overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If bicubic interpolation or super-resolution neural networks are used to upsample images, then image resolution is improved, but compression artifacts remain visible and computational demands increase
Solution Approach 1:
The patent segments the image processing task into multiple stages: (1) initial upsampling to increase resolution, (2) artifact detection to identify compression artifacts, and (3) selective refinement to remove artifacts while preserving image details. This segmentation allows the system to address both resolution enhancement and artifact reduction as separate but coordinated objectives.
Solution Approach 2:
The patent introduces an intermediary artifact detection mechanism that operates between the upsampling stage and the final output. This intermediary component identifies compression artifacts in the upsampled image and guides subsequent refinement operations, acting as a mediator that prevents artifacts from being amplified during the resolution enhancement process.
2Manufacturing precision
If super-resolution neural networks are used to reduce compression artifacts, then image quality is improved, but training and running times increase
Solution Approach 1:
The patent applies partial action by focusing computational resources only on regions containing compression artifacts rather than processing the entire image uniformly. The artifact detection mechanism identifies problematic areas, and refinement operations are applied selectively to these regions, reducing overall computational time while maintaining image quality.
Solution Approach 2:
The patent performs preliminary upsampling and artifact detection before applying refinement operations. By identifying artifact locations in advance through the detection mechanism, the system prepares a targeted refinement strategy that reduces computational overhead during the actual artifact removal process, thereby decreasing training and running times.
3Measurement precision
If image resolution is increased for printing or OCR tasks, then task accuracy is improved, but compression artifacts become more visible
Solution Approach 1:
The patent converts the harmful effect of compression artifacts into a beneficial process by using the artifact detection mechanism to identify and locate artifacts in the upsampled image. This detection capability transforms the previously harmful visible artifacts into actionable information that guides the refinement process, ultimately eliminating them and improving both print quality and OCR accuracy.
Solution Approach 2:
The patent applies local quality by performing refinement operations selectively on regions containing compression artifacts rather than uniformly across the entire image. The artifact detection mechanism identifies specific problematic areas, and refinement is applied locally to these regions, preserving image details in artifact-free areas while removing artifacts in problematic regions, thereby maintaining high task accuracy without amplifying artifacts.
Data Source
AI summary
An example system includes a feature extraction engine to identify features in a compressed image at a plurality of scales. The features include features corresponding to compression artifacts and features corresponding to image content to be upsampled. The feature extraction engine is to identify the features in the compressed image at a first scale based on noncontiguous pixels. The system also includes a reconstruction engine to refine the features corresponding to the image content to be upsampled and mitigate the features corresponding to the compression artifacts. The system includes an upsampling engine to generate an upsampled version of the compressed image based on the refined and mitigated features.