Image Processing Apparatus Feature Point Tracking for Camera-Shake Correction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image capturing systems face challenges in effectively tracking feature points corresponding to moving objects during the 'frame-in' period, leading to inadequate camera-shake correction, as they often track irrelevant feature points before the object enters the frame.
Innovation Solution
An image processing apparatus that sets a processing region and iterates through selecting and detecting tracking target points until a moving object is detected, then adjusts the selection process to follow the motion vectors of previously tracked points, ensuring accurate feature point tracking and camera-shake correction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If feature point tracking is started before the frame-in of a moving object, then the tracking process can begin early for camera-shake correction, but the system tracks irrelevant feature points instead of the moving object, reducing correction accuracy
Solution Approach 1:
The system performs preliminary feature point extraction and tracking preparation in advance, but delays the actual tracking execution until the moving object is detected. The processor extracts feature points from captured images and prepares tracking data structures beforehand, then activates tracking only when the moving object enters the frame, ensuring both early readiness and accurate tracking of the correct target
Solution Approach 2:
The system dynamically adjusts the tracking target selection based on the detection status of the moving object. Before frame-in, the system monitors for object appearance; after detection, it switches to tracking the moving object's feature points. This dynamic adaptation allows the system to optimize tracking accuracy according to the current capture phase while maintaining continuous operation
2Quantity of substance
If feature points are extracted from the entire image, then more feature points are available for tracking, but the distribution becomes non-uniform and processing complexity increases
Solution Approach 1:
The system applies different processing strategies to different regions of the image based on local characteristics. In regions containing the moving object, feature point extraction is performed with higher density and priority. In background regions, extraction is performed with lower density. This local differentiation ensures sufficient feature points for accurate tracking while reducing overall processing complexity and avoiding uniform distribution issues
3Productivity
If the tracking range is limited to improve processing efficiency, then fewer feature points need to be tracked, but the system may miss the moving object when it enters the frame
Solution Approach 1:
The system segments the image processing into distinct phases: a pre-frame-in phase where the entire image is monitored for object appearance, and a post-frame-in phase where tracking is focused on the detected moving object's region. This segmentation allows comprehensive monitoring during the critical detection period while enabling efficient localized tracking afterward, balancing efficiency and reliability
Data Source
AI summary
There is provided an image processing apparatus. A setting unit sets a first processing region in a part of a captured range. A selection unit selects a plurality of tracking target points in a first captured image of a plurality of captured images. A detection unit detects a motion vector of each tracking target point between the first captured image and a second captured image subsequently captured after the first captured image. A determination unit determines whether a moving object has appeared in the first processing region. A control unit performs control to iterate a first process set until it is determined that the moving object has appeared in the first processing region, and to iterate a second process set after it is determined that the moving object has appeared in the first processing region.