Single-Image Super-Resolution Flicker Suppression With DCT Detail Restoration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for single image super-resolution (SR) suffer from flickering artifacts and loss of details due to motion compensation errors, which propagate through video frames, leading to a degraded viewing experience.
Innovation Solution
A framework for flicker suppression that includes motion error correction and discrete cosine transform (DCT)-based detail restoration, aligning motion-compensated previous SR frames with current frames, and suppressing flickers on the frequency domain to generate high-resolution frames with enhanced details.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Stability of the object's composition
If motion compensation is applied to previous SR frames, then temporal coherence is improved, but motion compensation errors propagate and cause flickering artifacts
Solution Approach 1:
The patent extracts and separates the flickering component from the SR frame using frequency domain analysis. By identifying and isolating the flickering artifact in the frequency domain, the system can selectively remove it while preserving the underlying image content, thus resolving the contradiction between temporal coherence and flickering artifacts.
Solution Approach 2:
The patent converts the harmful flickering artifact into a beneficial signal for improvement. By detecting the flickering in the frequency domain and using it as a guide, the system applies targeted correction that actually enhances the temporal coherence while eliminating the harmful effects, turning the artifact into a useful reference for correction.
2Object-generated harmful factors
If motion compensation errors are corrected, then flickering is suppressed, but fine details may be lost
Solution Approach 1:
The patent applies local quality by treating different frequency components differently. In the frequency domain, it selectively processes high-frequency components (which contain fine details) separately from low-frequency components (which contain flickering artifacts). This allows targeted correction of flickering while preserving fine details through selective frequency band processing.
Solution Approach 2:
The patent transitions from spatial domain processing to frequency domain processing. By transforming the SR frame into the frequency domain using DCT or FFT, the system can correct flickering artifacts more effectively without losing fine details. The frequency domain provides an additional dimension where flickering and details are separated by frequency, enabling independent processing of each component.
3Productivity
If conventional SR methods are used, then processing speed is maintained, but flickering artifacts degrade image quality
Solution Approach 1:
The patent applies preliminary action by pre-processing the SR frame in the frequency domain before final output. The system performs frequency transformation, flickering detection, and correction as preliminary steps, then reconstructs the corrected SR frame. This preliminary frequency-domain processing enables efficient flickering suppression without significantly impacting overall processing speed, as the operations are optimized for computational efficiency.
Data Source
AI summary
One embodiment provides a method comprising receiving an input video comprising low-resolution (LR) frames and corresponding super-resolution (SR) frames, and generating a motion-compensated previous SR frame based on a current LR frame of the video and a motion-compensated previous residual frame of the video. The previous SR frame aligns with a current SR frame corresponding to the current LR frame. The method further comprises, in response to determining there is a mismatch between the previous SR frame and the current SR frame, correcting in the current SR frame errors that result from motion compensation based on the motion-compensated previous SR frame. The method further comprises restoring details to the current SR frame that were lost as a result of the correcting, and suppressing flickers of the current SR frame on the frequency domain, resulting in a flicker-suppressed current SR frame for presentation on a display.