Deep Learning Denoising and Super Resolution for Inspection Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inspection images generated using High Energy X-rays suffer from Poisson-Gaussian noise, making it difficult to detect objects of interest, such as threats or contraband, due to insufficient resolution, leading to time-consuming manual inspections.
Innovation Solution
A computer-implemented method using Deep Learning architectures for simultaneous denoising and super-resolution of inspection images, employing a synthetic data generator to train Deep Neural Networks, which enhance image resolution and remove noise without introducing artifacts, and a single Convolutional Neural Network for fast computation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If standard denoising methods are applied to HEX inspection images, then noise is reduced, but image resolution deteriorates and artifacts are introduced
Solution Approach 1:
The patent applies preliminary action by training the deep neural network on synthetic training data that simulates noisy low-resolution inspection images before actual denoising operations. This pre-training prepares the network to handle Poisson-Gaussian noise characteristics specific to HEX images, enabling it to perform denoising while preserving resolution and avoiding artifacts during actual application.
2Measurement precision
If manual inspection is performed to detect objects in noisy images, then detection accuracy improves, but inspection time increases
Solution Approach 1:
The patent replaces the mechanical manual inspection process with an automated deep learning-based image processing system. The trained neural network automatically denoises and enhances inspection images, providing detection capabilities that are both faster than manual inspection and capable of maintaining high accuracy by preserving fine details through the super-resolution component.
3Measurement precision
If image resolution is increased to improve object detection, then detection capability improves, but computation time increases
Solution Approach 1:
The patent applies preliminary action by pre-training the deep neural network on synthetic training data consisting of low-resolution images with simulated Poisson-Gaussian noise. This pre-training enables the network to learn effective denoising and super-resolution mappings, allowing it to rapidly process actual inspection images and produce high-resolution denoised outputs without requiring extensive computation during actual operation.
Data Source
AI summary
A computer-implemented method of processing one or more inspection images including a plurality of pixels includes obtaining an input inspection image generated by an inspection system configured to inspect one or more containers, wherein the inspection system is configured to inspect the container by transmission, through the container, of inspection radiation generated by an accelerator and having an angular divergence from the accelerator to an inspection radiation receiver including a plurality of detectors, the input inspection image having a higher noise, the higher noise including a Poisson-Gaussian noise whose variance is non-constant in the plurality of pixels, and a lower resolution; and processing the obtained input inspection image by applying, to the input inspection image, a trained machine learning algorithm for simultaneously increasing the lower resolution and decreasing the higher noise.