Temporal Differential Integration for Image Sensor Vibration Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for integrating stationary or quasi-stationary harmonic signals face challenges with irregular sampling, cut-off frequencies, and the inability to restore high-frequency noise, particularly in image sensors used in microscopes and satellites, leading to image blurring and geometric deformations.
Innovation Solution
A method that integrates vibration signals in temporal space using a linear combination of differentials, allowing for real-time correction of signals without preprocessing, and accounts for measurement noise, applicable to image sensors on carriers like satellites and microscopes, enabling local integration and restoration of main frequencies.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional Fourier transform methods are used to integrate harmonic signals, then the integration can be performed in frequency space, but the method produces cut-off frequencies at multiples of the integration step and cannot restore high-frequency components
Solution Approach 1:
Instead of transforming to frequency space and filtering (conventional approach), the patent inverts the approach by working directly in temporal space with differential measurements. The method uses temporal differentiation followed by linear combination to reconstruct the integrated signal, thereby preserving high-frequency components that would be lost in conventional frequency-space filtering.
Solution Approach 2:
The patent replaces the mathematical Fourier transform mechanism with a temporal differential mechanism. By taking time derivatives of the signal and forming linear combinations, the method achieves integration without requiring transformation to frequency space, thus avoiding the cut-off frequency problem inherent in conventional filtering approaches.
2Measurement precision
If polynomial modeling is used to integrate very low frequency noise, then the integration can be performed using least squares fitting, but the method is only valid for very low frequency signals and cannot handle quasi-stationary quasi-harmonic signals
Solution Approach 1:
The patent creates a universal method that handles multiple signal types (very low frequency noise, stationary harmonic signals, and quasi-stationary quasi-harmonic signals) through a single temporal differential framework. By using linear combinations of differential measurements with appropriately chosen coefficients, the method adapts to different signal characteristics without requiring separate processing approaches.
Solution Approach 2:
The patent changes the fundamental parameter from polynomial order (in low-frequency modeling) to differential measurement intervals and linear combination coefficients. This allows the same basic mechanism to effectively process signals across different frequency ranges and characteristics by adjusting the temporal sampling and weighting parameters rather than changing the underlying mathematical approach.
3Measurement precision
If conventional integration methods are applied after observing the whole signal, then the integration can be performed using all available time samples, but the integration cannot be performed locally or in real-time
Solution Approach 1:
The patent segments the signal processing into local temporal windows using differential measurements at specific time points. By forming linear combinations of differentials within localized time intervals, the method enables real-time integration without requiring observation of the entire signal beforehand, thus achieving both local processing and real-time capability.
Solution Approach 2:
The patent performs preliminary temporal differentiation of the signal to create a set of differential measurements that can be combined linearly to produce the integrated result. This preliminary transformation enables subsequent local integration operations to be performed independently and in real-time, as the differential relationships are established beforehand through the differentiation step.
4Reliability
If averaging of a large number of images is used to remove vibration, then the vibration can be reduced, but the method creates blurring of image edges
Solution Approach 1:
The patent introduces temporal differential measurements as an intermediary between the raw image signals and the final integrated result. By measuring and processing the differentials of the vibratory signal components separately, then combining them with appropriate coefficients, the method removes vibration effects without requiring simple averaging that blurs edges.
Solution Approach 2:
The patent replaces the mechanical averaging operation with a temporal differential and linear combination mechanism. Instead of averaging multiple images to reduce vibration (which blurs edges), the method uses time-derivative relationships to isolate and remove vibratory components while preserving the sharp edge information in the original images.
Data Source
Figure 1~3B
Figure 4~5
AI summary
The invention relates to an integration method for obtaining a harmonic or quasi-harmonic vibratory signal from at least one measure representative of a differential of said vibratory signal, said vibratory signal including n stationary or quasi-stationary known pulses ?i, wherein said method is characterised in that it comprises the following steps: an integrator integrates (S4) the vibratory signal formula (I) by carrying out a linear combination of m measures formula (II), for each time tk from a module (1), the linear combination having the form: formula (III) in which m = 2n, t k belongs to a time interval I including t; the measures are taken between the time t k and the times t k -tj distant by known time offsets tj; each coefficient w jk of the linear combination is a function of the known pulses ?i, of the known offsets tj, and of the measure position formula (II) in the time interval I. The invention also relates to a corresponding method for correcting a signal acquired by an image sensor.