Single Image Super-Resolution Using Transform-Invariant Directional Total Variation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image super-resolution methods face challenges in enhancing the resolution of single low-resolution images, particularly due to the limitations of existing techniques which often require multiple images or large training datasets, and struggle with preserving edge structures and reducing artifacts like jaggedness and blur.
Innovation Solution
A method utilizing affine transforms and Schattenp=1/2-norm and L1/2-norm penalties for decomposing images into low-rank and sparse components, followed by transform-invariant directional total variation regularization to generate high-resolution images from single low-resolution blurred images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional single-image super-resolution methods are used, then the resolution can be enhanced, but edge structures become blurred and artifacts like jaggedness appear
Solution Approach 1:
The patent segments the image into different components by applying affine transforms and decomposing into low-rank and sparse components. This segmentation allows independent optimization of edge structures (low-rank) and texture/details (sparse), preventing blur while reducing artifacts.
Solution Approach 2:
The patent changes the mathematical parameters by using Schattenp=1/2-norm and L1/2-norm penalties instead of traditional L1 or L2 norms. This parameter change enables more effective regularization that preserves edges while reducing artifacts during super-resolution reconstruction.
2Measurement precision
If multiple images are used for super-resolution, then resolution enhancement is improved, but the method becomes complex and requires large training datasets
Solution Approach 1:
The patent extracts the essential information from a single low-resolution image by applying affine transforms and decomposing into low-rank and sparse components. This extraction allows the method to achieve super-resolution performance previously requiring multiple images, simplifying the input requirement while maintaining high resolution enhancement.
Solution Approach 2:
The patent creates a universal framework that handles single-image super-resolution through affine transform invariance. The same low-rank and sparse decomposition framework works for various image types and resolutions, eliminating the need for multiple specialized methods or large training datasets.
3Manufacturing precision
If affine transforms are iteratively refined, then the decomposition accuracy improves, but the computational time increases
Solution Approach 1:
The patent implements iterative refinement of affine transforms with feedback mechanisms. The algorithm continuously adjusts the affine transforms based on the decomposition accuracy, stopping when convergence is achieved. This feedback loop ensures high decomposition accuracy while avoiding unnecessary computational iterations.
Data Source
AI summary
A super-resolution method for generating a high-resolution (HR) image from a low-resolution (LR) blurred image is provided. The method is based on a transform-invariant directional total variation (TI-DTV) approach with Schattenp=1/2 (S1/2-norm) and L1/2-norm penalties. The S1/2-norm and the L1/2-norm are used to induce a lower-rank component and a sparse component of the LR blurred image so as to determine an affine transform to be adopted in the TI-DTV approach. In particular, the affine transform is determined such that a weighted sum of the S1/2-norm and the L1/2-norm is substantially minimized. Based on the alternating direction method of multipliers (ADMM), an iterative algorithm is developed to determine the affine transform. The determined affine transform is used to transform a candidate HR image to a transformed image used in computing a directional total variation (DTV), which is involved in determining the HR image.