Optical Image Translation for Super-Resolution Imaging
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image processing methods in low light environments compromise image resolution due to large pixel sizes, leading to degradation in quality and increased noise, while complex architectures and high processing power are required to enhance resolution, making existing solutions costly and unreliable.
Innovation Solution
The system employs a time-varying linear phase mask (LPM) to generate multiple dissimilar low-resolution images, which are then interleaved using direction vectors to create an intermediate high-resolution image, compensating for pixel motion and generating a final high-resolution image without the need for complex architectures or high processing power.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Illumination intensity
If pixel size is increased to improve light sensitivity in low light environment, then light sensitivity is improved, but image resolution deteriorates due to reduction in pixel count
Solution Approach 1:
The patent applies segmentation by dividing the image capture process into multiple temporal segments. Multiple low-resolution images are captured at different time instances and then combined through super-resolution algorithms to reconstruct a high-resolution image, effectively segmenting the resolution enhancement task across time rather than requiring spatial segmentation through larger sensors
Solution Approach 2:
The patent transitions from spatial dimension to temporal dimension by capturing multiple images over time. Instead of increasing resolution through spatial arrangement of pixels (which would require larger sensors), the system uses temporal sequencing of multiple low-resolution captures to achieve high-resolution output, adding the time dimension to the imaging process
2Object-generated harmful factors
If exposure time is increased to collect more photons and reduce noise, then noise is reduced, but motion blur increases due to object motion within the exposure period
Solution Approach 1:
The patent employs periodic action by capturing multiple images at regular time intervals rather than using a single long exposure. This periodic sampling approach allows the system to collect sufficient photons across multiple short exposures while avoiding motion blur that would occur in a single long exposure, as each individual exposure is brief enough to freeze motion
Solution Approach 2:
The patent maintains continuity of useful action by continuously capturing images over a period of time rather than interrupting for a single long exposure. The continuous capture sequence ensures that motion is recorded across multiple frames, and super-resolution algorithms can select or combine frames to produce a sharp, high-resolution image without the motion artifacts of long exposure
3Measurement precision
If multiple cameras or moving cameras are used to capture images from different angles to enhance resolution, then image resolution is improved, but device complexity and cost increase
Solution Approach 1:
The patent uses copying by creating multiple virtual views of the same scene through computational methods rather than physical multiple cameras. Super-resolution algorithms generate high-resolution images by computationally synthesizing information from multiple temporal captures, effectively copying and combining data without requiring additional physical camera hardware
Solution Approach 2:
The patent replaces the mechanical system of multiple physical cameras or moving camera mechanisms with a computational system. Instead of mechanically moving cameras or using multiple camera bodies, the invention uses software-based super-resolution algorithms to achieve the same resolution enhancement, substituting mechanical complexity with computational processing
4Measurement precision
If super resolution methods such as dictionary learning are used to predict high resolution image from single low resolution image, then image resolution is improved, but computational intensity and memory requirements increase
Solution Approach 1:
The patent applies preliminary action by capturing multiple low-resolution images in advance at different time instances before the final high-resolution reconstruction is needed. This pre-capture of multiple frames provides the raw material for super-resolution algorithms, reducing the computational burden during final processing compared to methods that attempt to generate high-resolution images from single low-resolution inputs
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach effectively increases image resolution beyond the capabilities of the image sensor, improving light sensitivity and image quality without the drawbacks of existing methods, achieving up to 8× super-resolution with optimal color diversity and minimal computational intensity.
Implementation Method 1
a time varying linear phase mask (LPM) for generating multiple dissimilar low-resolution images
Data Source
AI summary
A method and system for enhancing image resolution using optical image translation, including: receiving multiple dissimilar low-resolution images generated by a time varying linear phase mask (LPM); receiving a direction vector for each of the multiple dissimilar low-resolution images; interleaving the received multiple dissimilar low-resolution images using the received direction vectors to form an intermediate high-resolution image; computing a likelihood of pixel motion in the intermediate high-resolution image; compensating for pixel motion in the intermediate high-resolution image; and generating a final high-resolution image.