Optical-Aware Image Sharpening for Aberration-Specific Blur
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image sharpening methods using Wiener filters and CNNs struggle to accurately sharpen blurred images caused by various aberrations in optical systems while managing high learning loads and data storage requirements.
Innovation Solution
An image processing method utilizing a machine learning model, such as a CNN, that incorporates optical system information to sharpen or reshape blurs by learning weights specific to each optical system state, reducing the learning load and data storage needs.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Wiener filter-based linear processing is used for sharpening, then the processing is simple and fast, but the sharpening accuracy is insufficient and cannot restore information where spatial frequency spectrum is zero
Solution Approach 1:
The patent replaces the traditional Wiener filter-based linear processing system with a deep neural network-based nonlinear processing system. The DNN learns optimal sharpening operations through training data, enabling it to restore image information that linear methods cannot recover, particularly where spatial frequency spectrum is zero or where blur characteristics are complex.
2Measurement precision
If different Wiener filters are used for different aberrations, then the sharpening accuracy for each aberration type improves, but the amount of stored data increases significantly
Solution Approach 1:
The patent employs a single deep neural network that can handle multiple types of aberrations (defocus, spherical aberration, coma, astigmatism, etc.) simultaneously. The DNN is trained on diverse training data representing various aberration types and conditions, enabling one universal model to perform sharpening across all aberration scenarios without requiring separate filters for each type.
Solution Approach 2:
The patent changes the approach from storing multiple discrete filter parameters to learning continuous parameter transformations through the DNN. The network learns to adapt its weights and biases based on input image characteristics and aberration types, dynamically adjusting the sharpening operation without requiring pre-stored filter sets for each condition.
3Productivity
If the CNN learns all blurs using averaging, then the learning load is reduced, but the sharpening accuracy for each specific blur type decreases
Solution Approach 1:
The patent applies local quality by training the DNN to recognize and process different blur types with specialized features. The network learns distinct patterns for different aberrations through diverse training data, enabling it to apply appropriate sharpening strategies for each local blur characteristic rather than using a single averaged approach. This allows high accuracy for each specific blur type while maintaining reasonable learning efficiency.
4Measurement precision
If blurs are divided into multiple groups for individual learning, then the sharpening accuracy for each group improves, but the learning load and amount of stored data increase significantly
Solution Approach 1:
The patent merges the learning of multiple blur types into a single unified deep neural network model. Instead of creating separate models for different blur groups, the DNN is trained on comprehensive data covering all aberration types simultaneously. The network architecture and training process are designed to learn shared and specific features across different blur types, achieving high accuracy for each type while avoiding the computational burden of multiple separate models.
Data Source
AI summary
An image processing method includes a first step of acquiring input data including a captured image and optical system information relating to a state of an optical system used for capturing the captured image and a second step of inputting the input data to a machine learning model and of generating an estimated image acquired by sharpening the captured image or by reshaping blurs included in the captured image.