Subpixel Registration for Super-Resolution Image Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image registration techniques, such as pixel-level registration, often result in artifacts like blurring and double features when constructing super-resolution images, due to limitations in pixel size and resolution, which can increase fabrication costs and noise.
Innovation Solution
The method involves sub-pixel image alignment by mapping pixels from one image to a coordinate system, applying sub-pixel shifts, calculating energy based on gradients, and determining the shift that minimizes overall energy to achieve precise alignment and reduce artifacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If pixel-level registration is applied to construct super-resolution images, then image resolution can be improved, but artifacts such as blurring and double features appear
Solution Approach 1:
The patent applies sub-pixel level shifts (changes in positional parameters) to align images more precisely than traditional pixel-level registration. By operating at a finer granularity (sub-pixel rather than full-pixel shifts), the method achieves better alignment accuracy, reducing registration errors that cause blurring and double features in super-resolution images
Solution Approach 2:
The patent replaces traditional mechanical/optical image processing with computational energy minimization. Instead of relying on physical alignment mechanisms or simple pixel-matching algorithms, the system uses an energy function that evaluates gradient differences across multiple images and iteratively adjusts sub-pixel shifts to minimize overall energy, thereby eliminating artifacts through mathematical optimization rather than mechanical means
2Measurement precision
If higher resolution sensors are used to obtain super-resolution images, then image quality improves, but fabrication costs and shot noise increase
Solution Approach 1:
The patent creates a high-resolution image by computationally combining multiple lower-resolution images rather than using a single high-resolution sensor. Each low-resolution image serves as a copy that, when properly aligned and integrated through energy minimization, contributes to reconstructing the super-resolution image, avoiding the need for expensive high-resolution sensor fabrication
Solution Approach 2:
The patent merges multiple lower-resolution images into a single super-resolution image through computational processing. By combining information from multiple images taken at different sub-pixel offsets, the system synthesizes higher resolution detail that would otherwise require a more expensive sensor, effectively merging the capabilities of several sensors into one virtual high-resolution sensor
3Measurement precision
If sensor chip size is increased to accommodate more pixels, then resolution improves, but capacitance increases and data transfer rate decreases
Solution Approach 1:
The patent uses multiple copies of lower-resolution images to achieve the resolution of a larger sensor. Instead of physically increasing sensor size and dealing with associated capacitance and bandwidth limitations, the system processes multiple smaller image copies computationally, maintaining data transfer efficiency while achieving high resolution through post-processing
Solution Approach 2:
The patent transitions from a spatial dimension problem (larger sensor chip) to a temporal/computational dimension solution. Rather than increasing physical sensor dimensions which worsens capacitance and bandwidth issues, the system uses time-multiplexed acquisition of multiple images and applies computational algorithms to achieve the same resolution goal without the physical constraints
4Ease of operation
If pixel-level registration is used, then alignment can be achieved, but sub-pixel precision is lost resulting in registration errors
Solution Approach 1:
The patent changes the precision level of the alignment parameter from pixel-level (integer shifts) to sub-pixel level (fractional shifts). By allowing shift values with decimal precision rather than whole-pixel increments, the system achieves finer alignment accuracy, reducing registration errors that would otherwise be inherent in coarse pixel-level registration
Solution Approach 2:
The patent implements an iterative feedback mechanism where the energy function evaluates the current alignment quality and provides feedback to adjust sub-pixel shifts. The system repeatedly calculates energy based on gradient differences, determines whether the minimum has been reached, and continues adjusting shifts accordingly, creating a closed-loop feedback system that progressively improves registration precision beyond what fixed pixel-level alignment can achieve
Data Source
AI summary
Sub pixel image alignment includes mapping first pixels from a first image and second pixels from a second image to a coordinate system and applying one or more sub-pixel shifts to the mapped first pixels. For each sub-pixel shift, an overall energy is calculated and is based on a plurality of gradients that represent changes in a channel value among the shifted first pixels and the mapped second pixels. The sub-pixel alignment further includes determining the sub-pixel shift that provides the lowest overall energy.