Camera Shake Correction Using Motion Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional imaging devices face challenges in accurately correcting camera shake, especially when capturing night scenes or images of plain walls, due to difficulties in extracting feature points, which decreases the precision of camera shake correction.
Innovation Solution
A motion information obtaining device and method that includes an image inputter, a detector to identify purposeful motion, and an obtainer to extract motion information other than purposeful motion, allowing for improved camera shake correction by distinguishing between intended and unintended motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If feature point extraction is used for camera shake correction, then camera shake correction can be performed, but detection precision decreases when imaging night scenes or plain walls
Solution Approach 1:
The patent introduces an intermediary classification process that categorizes motion into purposeful and non-purposeful types. This intermediary step acts as a mediator between the raw motion detection and the camera shake correction, allowing the system to handle different motion types appropriately and maintain precision in challenging imaging conditions.
Solution Approach 2:
The patent segments the motion detection task into distinct components: detecting overall motion, classifying motion type (purposeful vs. non-purposeful), and selectively applying correction. This segmentation allows the system to process different types of motion differently, improving precision by excluding purposeful motion from camera shake correction.
2Measurement precision
If motion vectors are calculated from feature points, then camera shake amount can be estimated, but precision decreases when feature points cannot be extracted
Solution Approach 1:
The classification of motion into purposeful and non-purposeful types serves as an intermediary that bridges the gap between difficult feature point extraction and accurate camera shake measurement. By filtering out purposeful motion through this intermediary classification, the system can achieve precise camera shake estimation even when feature points are hard to extract.
Solution Approach 2:
The patent extracts and removes the purposeful motion component from the overall motion detection. By taking out the purposeful motion element, the system can focus on measuring only the non-purposeful camera shake, thereby improving measurement precision in conditions where feature point extraction is difficult.
3Device complexity
If all detected motion is used for correction, then simple correction process is maintained, but precision decreases due to inclusion of purposeful motion
Solution Approach 1:
The patent segments the motion correction process into distinct stages: motion detection, motion classification (purposeful vs. non-purposeful), and selective correction application. This segmentation adds minimal complexity while dramatically improving precision by ensuring that only non-purposeful motion is used for camera shake correction.
Solution Approach 2:
The system dynamically adjusts the correction process by classifying motion in real-time. Based on the classification result, the system adaptively determines whether to apply correction, making the correction process dynamic and context-dependent rather than static, thereby improving precision without excessive complexity.
Data Source
AI summary
A motion detector calculates a camera shake matrix H(t) applied for a coordinate conversion for a camera shake correction. At this time, the motion detector calculates a motion vector (MV) for each block obtained by dividing a frame, and excludes an object motion vector (MV-B) from the MV. In addition, when a camera shake matrix H_mv that represents a camera shake between the frames is not calculable based on the MV, the motion detector calculates and adjusts a camera shake matrix H_sensor that represents a camera shake in accordance with a motion of an image picker, and settles this matrix as the camera shake matrix H(t).