Super-Resolution Network Face Integrity via Matching Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Super-resolution networks using GANs generate high-frequency components not present in the input signal, leading to deviations in human faces, such as slight shifts in eye and mouth shapes, resulting in changed appearances during image upscaling.
Innovation Solution
An information processing device comprising a human face determination network to calculate the matching degree between input and output images, and a super-resolution network that adjusts its generation force based on this matching degree to minimize changes in human faces during processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the generation force of super-resolution processing is increased to generate high-frequency components, then image resolution and detail quality are improved, but deviation from the input image occurs leading to changes in human face appearance
Solution Approach 1:
The patent implements a feedback mechanism where the calculated human face matching degree is fed back to dynamically adjust the generation force of the super-resolution network. When the matching degree indicates potential face distortion, the generation force is reduced to prevent appearance changes, thus resolving the contradiction between resolution enhancement and face consistency.
Solution Approach 2:
The patent makes the generation force dynamic by adjusting it based on the calculated human face matching degree. Instead of using a fixed generation force, the system adaptively modulates it to maintain face appearance consistency while still achieving high-resolution output when appropriate, thereby resolving the static contradiction between resolution and consistency.
2Manufacturing precision
If signal generation capability is enhanced to produce high-resolution images, then image quality is improved, but deviation from input signal occurs causing face shape changes
Solution Approach 1:
The system uses feedback from the human face matching degree calculation to control the signal generation process. When the matching degree suggests that face shape fidelity is at risk, the generation force is adjusted downward, preventing loss of critical face shape information while still achieving high image quality through controlled generation.
Solution Approach 2:
The patent changes the parameter of generation force dynamically based on the human face matching degree. By adjusting this parameter in response to calculated matching results, the system maintains face shape fidelity while achieving high image quality, resolving the contradiction between these two requirements.
Data Source
AI summary
The information processing device (IP) includes a human face determination network (PN) and a super-resolution network (SRN). The human face determination network (PN) calculates a human face matching degree between an input image (IMI) before being subjected to super-resolution processing and an input image (IMI) after being subjected to the super-resolution processing. The super-resolution network (SRN) adjusts a generation force of the super-resolution processing based on the human face matching degree.