Motion Vector Calculation for Wide Dynamic Range Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing wide dynamic range (WDR) technologies face inaccuracies in motion vector detection for blur correction when combining long exposure and short exposure images with different exposure levels, leading to incomplete blur correction in certain image regions.
Innovation Solution
An apparatus and method that calculate motion vectors for long and short exposure images by extracting feature points based on saturation degree and edge strength, dividing regions into sub-regions, and obtaining local and global motion vectors through block matching and gamma correction to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If WDR synthesis is performed first to combine long exposure and short exposure images, then a synthesized image is obtained, but the synthesized image may not be uniform for each synthesized region due to position differences between frames with different exposure levels
Solution Approach 1:
The patent divides the image into multiple regions and performs separate motion vector calculations for each region. This segmentation allows the system to handle position differences between frames with different exposure levels on a regional basis, improving the uniformity of the synthesized image while maintaining accurate motion vector detection for each specific region.
2Measurement precision
If blur correction is performed with long exposure and short exposure image frames with different exposure levels prior to WDR synthesis, then motion vector detection may be conducted, but the frames with different exposure levels may have very different luminance components, such that motion vector detection for blur correction may be very inaccurate
Solution Approach 1:
The patent applies different processing strategies to different regions of the image based on their exposure characteristics. By performing motion vector detection separately for long exposure regions and short exposure regions, the system accounts for the different luminance components in each region, thereby improving motion vector detection accuracy despite the overall luminance differences between frames.
Data Source
AI summary
Provided an apparatus for obtaining a motion vector including at least one processor configured to receive a long exposure image and a short exposure image of the current frame, and obtain a motion vector of the long exposure image and a motion vector of the short exposure image, the processor being configured to extract feature points in the long exposure image based on a saturation degree of the long exposure image, and obtain a motion vector of the long exposure image based on a difference from a long exposure image of a previous frame, and the processor being configured to extract feature points in the short exposure image and obtain a motion vector of the short exposure image based on a difference from a short exposure image of a previous frame based on a saturation degree of the short exposure image.