Discrete Resolution Models for Real-Time Image Restoration
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing high-resolution image restoration technologies, particularly continuous image super-resolution methods, are limited in their applicability to electro-optical equipment due to complex computation requirements and are inefficient for real-time processing, especially when high magnification is needed.
Innovation Solution
An image restoration method and apparatus that generate independent restoration models for different resolutions through learning, allowing for the selection and application of a suitable restoration model based on the resolution of a distorted image to restore it into an improved upscaled image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If continuous image super-resolution restoration technology is used, then image restoration quality is improved, but processing speed deteriorates and real-time processing becomes difficult
Solution Approach 1:
The patent divides the continuous image super-resolution process into discrete resolution levels (e.g., 2x, 3x, 4x magnification). Separate restoration models are trained for each discrete resolution level, transforming the continuous problem into segmented discrete problems that can be processed more efficiently in real-time applications.
Solution Approach 2:
The patent pre-trains multiple restoration models for different resolution levels before actual image restoration is needed. These pre-trained models are stored and can be quickly selected and applied during real-time processing, eliminating the need for complex computations during the actual restoration operation.
2Speed
If single image super-resolution restoration technology is used, then processing speed is improved, but image restoration quality deteriorates due to insufficient information
Solution Approach 1:
The patent creates a universal framework that handles both single-image and continuous-image super-resolution scenarios. The system can select appropriate restoration models based on input characteristics, making it versatile for different application scenarios while maintaining both speed and quality.
3Adaptability or versatility
If existing super-resolution restoration technology is applied to electro-optical equipment, then image magnification capability is improved, but device complexity and computation requirements increase beyond available resources
Solution Approach 1:
The patent changes the parameter of resolution magnification into discrete levels (2x, 3x, 4x). Each level has a dedicated optimized model with appropriate computational complexity for that specific magnification level, allowing electro-optical equipment to achieve high magnification capabilities without requiring a single overly complex system.
Data Source
AI summary
The present embodiment provides an image restoration method and apparatus which generate independent different restoration models by performing learning for each of different resolutions, receive a distorted image, and apply a restoration model corresponding to the resolution of the distorted image among the independent different restoration models to restore the distorted image into an improved upscaled image centering on a restoration target object within the distorted image.