Motion Vector Acquisition for Camera Shake Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems struggle to accurately estimate blur amounts due to radial motion vectors when an imaging device is mounted on a movable body, as these vectors interfere with the estimation of camera shake and alignment between frames.
Innovation Solution
An image processing device that acquires motion vectors and adjusts the calculation area based on the orientation and movement situation of the imaging device, excluding areas prone to radial motion vectors by using sensors for orientation and movement data to set a motion vector calculation exception area, thereby enhancing the accuracy of blur estimation and alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If motion vectors are acquired from all image areas, then the quantity of motion vector data increases, but the accuracy of blur estimation deteriorates due to radial motion vectors interfering with camera shake detection
Solution Approach 1:
The image is divided into multiple regions, and motion vectors are selectively acquired only from specific regions that are less affected by radial motion vectors. This segmentation allows the system to maintain sufficient motion vector data quantity while improving blur estimation accuracy by excluding problematic areas.
Solution Approach 2:
Different regions of the image are treated differently based on their susceptibility to radial motion vectors. The system identifies and selects regions with lower radial motion vector interference for motion vector acquisition, applying local quality assessment to optimize both data quantity and estimation accuracy.
2Adaptability or versatility
If the image capture device is mounted on a movable body to enable flexible imaging, then the adaptability of the system improves, but radial motion vectors are generated that interfere with blur estimation
Solution Approach 1:
The system extracts and removes the harmful radial motion vector components from the motion vector data by selectively acquiring motion vectors only from regions less affected by radial motion. This extraction approach maintains the adaptability of mounting on movable bodies while eliminating the interfering radial motion effects.
Solution Approach 2:
The system converts the harmful effect of radial motion vectors into a benefit by using the known orientation information to identify and exclude regions affected by radial motion, thereby transforming the mounting flexibility advantage into accurate blur estimation despite the presence of radial motion vectors.
3Area of stationary object
If motion vectors are acquired from peripheral image areas, then the coverage of motion vector data increases, but the accuracy of camera shake detection deteriorates due to radial motion vector interference
Solution Approach 1:
The image area is segmented into regions with different characteristics regarding radial motion vector susceptibility. Motion vectors are acquired from selected regions that provide sufficient coverage while avoiding areas where radial motion vectors would significantly degrade camera shake detection accuracy.
Data Source
AI summary
By using various information such as GPS information, cartographic information, and atmospheric pressure acquired by a sensor, a CPU of an imaging device calculates a motion vector in a limited MV calculation area acquired by excluding from images an area where radial motion vectors occur which highly possibly give an adverse effect to motion vector calculation, in accordance with the gravity direction with respect to the imaging device, the orientation of the imaging device, and the slope angle of the road where the imaging device is moving forward. After calculating the motion vector, the CPU estimates a blur amount between frames from the calculated motion vector and, based on this blur amount, performs the alignment of a feature point between the frames by image deformation processing.