Bayesian Super Resolution Image Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for obtaining higher resolution images from lower resolution images face challenges due to the ill-posed nature of inverse mathematical problems, where existing signal processing techniques often fail to accurately handle noise and outliers, leading to suboptimal results in image reconstruction.
Innovation Solution
A Bayesian estimation method using robust, non-Gaussian noise modeling to improve the accuracy of super resolution image reconstruction by downplaying the impact of outliers and employing robust probability functions such as the Huber or Tukey functions, along with anisotropic diffusion filters, to iteratively refine the image alignment and noise estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If robust non-Gaussian noise modeling is used to handle outliers in image reconstruction, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent changes the noise model parameters from standard Gaussian distribution to robust non-Gaussian distributions (such as Huber or Tukey functions). This parameter change allows the system to better handle outliers and impulsive noise in the image reconstruction process, improving measurement precision while managing the increased computational complexity through efficient algorithmic implementation
Solution Approach 2:
The patent introduces robust probability functions as intermediary elements between the observed low-resolution images and the reconstructed high-resolution image. These functions act as mediators that filter out the influence of outliers and noise, allowing accurate reconstruction without directly processing the problematic noisy data, thus improving precision while containing complexity
2Manufacturing precision
If multiple lower resolution images are processed to obtain higher resolution image, then manufacturing precision is improved, but loss of time increases
Solution Approach 1:
The patent performs preliminary alignment and registration of multiple low-resolution images before the main reconstruction process. By pre-processing the images to establish accurate geometric relationships and reduce misalignment errors, the system improves the final reconstruction quality while reducing the computational burden and time required for the subsequent high-resolution synthesis
Solution Approach 2:
The patent divides the image reconstruction process into multiple stages: alignment, noise modeling, and reconstruction. This segmentation allows each stage to be optimized independently, with parallel processing capabilities in the alignment and noise estimation phases, thereby improving overall processing efficiency while maintaining high reconstruction quality
Data Source
AI summary
A result higher resolution (HR) image of a scene given multiple, observed lower resolution (LR) images of the scene is computed using a Bayesian estimation image reconstruction methodology. The methodology yields the result HR image based on a Likelihood probability function that implements a model for the formation of LR images in the presence of noise. This noise is modeled by a probabilistic, non-Gaussian, robust function. The image reconstruction methodology may be used to enhance the image quality of images or video captured using a low resolution image capture device. Other embodiments are also described and claimed.