Image Shake Correction Using Geometric Deformation and Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image stabilization methods in image capture apparatuses fail to accurately correct image shakes caused by six-degree-of-freedom movements, particularly due to difficulties in detecting and estimating motion vectors, especially in low-contrast scenes and when correcting rolling shutter distortions, leading to potential errors and increased computational complexity.
Innovation Solution
An image capture apparatus and control method that utilize a detection unit to calculate shake information, a calculation unit to determine shake correction amounts, and a deformation unit to geometrically deform images, canceling changes caused by rotational and exposure differences, thereby accurately correcting image shakes in directions perpendicular to the optical axis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If geometric deformation processing is applied using motion vectors for six-degree-of-freedom movement correction, then image shake correction accuracy is improved, but computational complexity increases explosively and estimation becomes unstable
Solution Approach 1:
The patent segments the six-degree-of-freedom motion correction into two parts: (1) rotational motion correction using sensor data for yaw and pitch axes, and (2) translational motion correction using geometric deformation processing only for roll axis. This segmentation avoids the computational explosion of estimating all six parameters simultaneously while maintaining correction accuracy for the translational components that sensors cannot detect.
Solution Approach 2:
The patent extracts and isolates the roll axis rotational motion as the specific component requiring geometric deformation processing. By identifying that conventional sensors cannot detect roll axis rotation and translational motion, the invention applies motion vector-based correction only to this specific component rather than attempting to correct all six degrees of freedom, thereby reducing computational complexity while maintaining necessary correction accuracy.
2Adaptability or versatility
If motion vectors are used to detect image distortion for rolling shutter correction, then correction capability is improved, but detection accuracy decreases in low-contrast scenes leading to erroneous vectors
Solution Approach 1:
The patent introduces an intermediary approach by using sensor-detected shake information as a reference to guide and constrain the geometric deformation processing. Instead of relying solely on motion vectors from low-contrast images, the sensor data serves as a mediator that provides reliable rotational motion information, allowing the geometric deformation to focus only on estimating the roll component and translational motion, thereby improving detection accuracy in challenging scenes.
3Reliability
If sensors detect rotational motion around yaw and pitch axes, then optical anti-shake function is achieved, but translational motion and roll axis rotation remain undetected
Solution Approach 1:
The patent replaces the mechanical sensor-based detection system with a computational approach using geometric deformation processing. Instead of adding physical sensors to detect translational motion and roll axis rotation (which would increase device complexity and cost), the invention uses image-based geometric deformation algorithms to estimate and correct for these undetected motion components, substituting mechanical detection with computational analysis.
Data Source
AI summary
An amount of change between images is calculated based on a motion vector between images obtained by an image sensor and on shake information related to a shake in a rotational direction around at least an axis perpendicular to an optical axis of an imaging optical system. Then, based on a shake correction amount determined based on the shake information and the amount of change, an image obtained by the image sensor is geometrically deformed. The amount of change is a shake component of the image capture apparatus in a direction perpendicular to the optical axis, and is calculated after cancelling at least one of a change between images caused by the shake in the rotational direction included in the shake information and a change between images caused by an influence of a difference between exposure periods of the image sensor.