Image Stabilization Feedback for Sensor Drift and Translational Shake
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image stabilization systems using angular velocity sensors face challenges due to offset calculation errors and sensitivity variations, leading to incomplete image stabilization, especially in macro imaging and panning shots, where translational shake components are significant and accurate correction is difficult to achieve.
Innovation Solution
An image stabilization apparatus and method that identifies and corrects for angular velocity sensor output variations using a combination of angular velocity signals, motion vectors, and position signals, converting these into effective shake correction amounts to improve image stabilization performance by accounting for sensor offset and sensitivity fluctuations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If an angular velocity sensor is used for shake detection, then real-time shake detection is enabled, but offset calculation errors and sensitivity variations occur leading to insufficient stabilization precision
Solution Approach 1:
The patent implements feedback by detecting actual image shake through motion vectors and comparing it with sensor-based shake detection results. The difference between these measurements is fed back to correct the angular velocity sensor output, thereby compensating for offset errors and sensitivity variations while maintaining real-time detection capability
Solution Approach 2:
The patent substitutes the purely sensor-based mechanical detection system with a hybrid approach that uses image processing (motion vector analysis) to verify and correct sensor measurements. This replaces reliance on the angular velocity sensor alone with a combination of sensor data and visual evidence from the captured images
2Reliability
If pseudo integral is applied in low-pass filter for integration processing, then offset noise is reduced, but ultra-low-frequency component performance is sacrificed
Solution Approach 1:
The patent segments the shake correction into different frequency components by separately processing high-frequency noise through the low-pass filter with pseudo integral, while preserving ultra-low-frequency components through motion vector-based correction. This segmentation allows different processing strategies for different frequency ranges
3Productivity
If angular velocity sensor offset is updated in real time by feedback control, then prediction convergence is improved, but anti-shake performance deteriorates during prediction
Solution Approach 1:
The patent performs preliminary correction by calculating motion vectors from image frames before applying them to correct the shake. This preliminary action based on actual image content provides a reliable correction foundation that is not affected by real-time offset updates, thereby maintaining anti-shake performance stability while enabling fast convergence
Data Source
AI summary
An image stabilization apparatus comprises: an identification unit that, based on an angular velocity signal outputted from a shake detection unit, a motion vector calculated from a difference between frames of images outputted from an image sensor, and a position signal indicating a position of a correction unit that corrects a shake optically, identifies an output variation of the shake detection unit; and a conversion unit that converts the angular velocity signal into a shake correction amount by correcting the angular velocity signal based on the output variation identified by the identification unit.