Image Stabilization via Zoom-Adaptive Motion Vector Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies fail to effectively stabilize images captured by devices that experience external shocks or have moving objects, especially at high zoom magnifications, leading to image distortion and instability.
Innovation Solution
An image processing apparatus and method that includes a feature point extractor, motion vector extractors, and an image stabilizer, which detect feature points, extract local and global motion vectors, and correct image shaking by selecting effective motion vectors based on zoom magnifications and sensor data, thereby stabilizing images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If image stabilization is applied using conventional motion vector extraction methods, then image shaking can be corrected under normal conditions, but image distortion occurs when external shocks are applied or moving objects are present in the captured image
Solution Approach 1:
The patent applies different algorithms for selecting effective local motion vectors based on zoom magnification levels. At high zoom magnifications, a motion area-based selection algorithm is used, while at normal zoom levels, a sensor data-based algorithm is applied. This local differentiation resolves the contradiction by optimizing the stabilization approach for specific operating conditions, preventing both shaking and distortion simultaneously.
Solution Approach 2:
The patent changes the selection criteria for effective local motion vectors based on the zoom magnification parameter. When zoom magnification exceeds a predetermined threshold, the system switches from sensor data-based selection to motion area-based selection. This parameter-driven adaptation allows the system to maintain image quality across varying zoom conditions while effectively stabilizing the image.
2Reliability
If motion vectors from all areas are used for stabilization, then comprehensive motion coverage is achieved, but moving objects cause incorrect stabilization and image distortion
Solution Approach 1:
The patent extracts and excludes motion vectors from moving objects by identifying motion areas in the image. By detecting regions with significant motion discrepancies and removing their corresponding motion vectors from the stabilization calculation, the system achieves accurate stabilization of static scenes while ignoring moving objects, thus preventing distortion.
Solution Approach 2:
The patent implements a feedback mechanism where motion areas are detected based on motion vector analysis, and this detection feeds back into the selection of effective local motion vectors. The system continuously refines its understanding of the scene by using motion area information to filter out unreliable motion vectors, creating a closed-loop stabilization process that maintains accuracy.
3Measurement precision
If high zoom magnification is used to capture distant objects, then object detail is improved, but image shaking and distortion become more pronounced
Solution Approach 1:
The patent applies a specialized motion vector selection algorithm specifically for high zoom magnification conditions. When the zoom magnification exceeds a predetermined value, the system uses motion area-based effective local motion vector selection, which is optimized for the challenges of high-zoom stabilization, thereby maintaining both detail resolution and image stability.
Solution Approach 2:
The patent performs preliminary detection of motion areas before finalizing the stabilization process. By identifying motion areas in advance and using this information to select effective local motion vectors, the system prepares the stabilization calculation with accurate motion data, preventing shaking and distortion even at high zoom magnifications where these effects are magnified.
Data Source
AI summary
Provided is an image processing apparatus including at least one processor configured to implement a feature point extractor that detects a plurality of feature points in a first image input from an image sensor; a first motion vector extractor that extracts local motion vectors of the plurality of feature points and select effective local motion vectors from among the local motion vectors by applying different algorithms according to zoom magnifications of the image sensor; a second motion vector extractor that extracts a global motion vector by using the effective local motion vectors; and an image stabilizer configured to correct shaking of the first image based on the global motion vector.