Kalman Filter Image Blurring Correction for Camera Shake
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image blurring correction methods for camera shaking suffer from characteristic losses due to bandpass filtering, which reduces gain and causes phase fluctuations, leading to inaccurate shake correction, especially in the frequency band close to camera shaking frequencies.
Innovation Solution
The implementation of a model-based filtering process using a Kalman filter to estimate translational camera shaking characteristics, reducing errors and characteristic losses by modeling user-specific camera shaking patterns and using auto-regressive models to select optimal filtering parameters.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If bandpass filtering is applied to correct image blurring, then the correction process can be implemented, but characteristic losses occur including gain reduction and phase fluctuations leading to inaccurate shake correction
Solution Approach 1:
The patent changes the filtering approach from bandpass filtering to Kalman filtering, which uses a different set of parameters (state transition models, observation models, and covariance matrices) to achieve frequency-dependent correction without the characteristic losses of traditional bandpass filters. This allows accurate shake correction across different frequency bands while maintaining gain and phase integrity.
Solution Approach 2:
The patent replaces the mechanical bandpass filtering system with a model-based Kalman filtering system that uses auto-regressive models to represent camera shaking characteristics. This substitution eliminates the inherent limitations of bandpass filters in the frequency domain while maintaining the ability to correct image blurring effectively.
2Measurement precision
If conventional bandpass filtering is used for shake correction, then the system structure remains simple, but accuracy deteriorates due to gain reduction and phase fluctuations
Solution Approach 1:
The patent substitutes the simple but inaccurate bandpass filter with a more complex Kalman filter based system that uses auto-regressive models. While the computational complexity increases, this substitution provides significantly improved accuracy by modeling the temporal characteristics of camera shaking rather than relying on fixed frequency band filtering.
Solution Approach 2:
The patent introduces dynamic modeling through auto-regressive models that adapt to the temporal characteristics of camera shaking. Instead of using static bandpass filters with fixed frequency responses, the system dynamically models the shaking behavior over time, allowing for more accurate correction while managing complexity through efficient recursive calculation.
Data Source
AI summary
An image blurring correction apparatus includes: an optical system that forms a subject image on an image formation plane; an acceleration sensor that detects an acceleration of an apparatus that has the image formation plane; a first calculation circuit that calculates a first state quantity by using the acceleration; a second calculation circuit that calculates, as a second state quantity, an estimate value obtained by estimating a true value of the first state quantity by using the first state quantity and a model; and a third calculation circuit that calculates an amount of image blurring on the image formation plane by using the second state quantity.