Fisheye Image Subject Distortion Correction via Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Fisheye images taken with a fisheye lens suffer from significant distortion, making it difficult for users to clearly identify and confirm subjects, especially at the peripheral parts of the image, due to the increasing distortion from the center to the edges.
Innovation Solution
An image processing device and method that recognizes predetermined subjects in wide-angle images, corrects the distortion of the subject's image, and associates the corrected image with the original wide-angle image, allowing for easier subject identification and visualization by displaying the corrected image alongside face images of individuals.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of moving object
If a fisheye lens is used to capture a wide-angle image, then the angle of view is substantially 180° and a wide-range image is obtained, but significant distortion occurs from the center to the peripheral parts of the image
Solution Approach 1:
The patent divides the wide-angle image into multiple regions (center region and peripheral regions) and applies different processing to each region. Face recognition is performed on the center region with minimal distortion, while the peripheral regions are processed to associate corrected face images with the distorted wide-angle view, allowing users to identify subjects throughout the entire image area despite the distortion.
Solution Approach 2:
The patent introduces corrected face images as an intermediary element that bridges the gap between the distorted wide-angle image and the user's need to clearly identify subjects. These corrected images, extracted and processed separately, serve as a reference that helps users understand who is present in different parts of the distorted wide-angle image without directly altering the original wide-angle view.
2Measurement precision
If the entire wide-angle image is corrected for distortion, then subject identification is improved, but the correspondence relationship with the original fisheye image becomes unclear
Solution Approach 1:
The patent separates the subject identification function from the wide-angle image display by extracting face images and correcting their distortion independently. This segmentation allows the original distorted wide-angle image to remain unchanged (preserving the correspondence relationship) while providing separately corrected face images for clear subject identification.
Solution Approach 2:
The patent creates corrected copies of face images from the distorted wide-angle image. These copies are processed to remove distortion while the original wide-angle image remains intact. Users can view both the original distorted image and the corrected face image copies, maintaining the correspondence relationship while enabling clear subject identification.
3Measurement precision
If face images are extracted and corrected separately, then subject identification is enhanced, but the device complexity increases
Solution Approach 1:
The patent performs face recognition and image extraction as preliminary actions on the center region of the wide-angle image where distortion is minimal. By completing these processing steps before final display and using the extracted face information to guide peripheral region processing, the system reduces the overall complexity compared to processing the entire distorted image.
Data Source
AI summary
A control unit 21 of a main device 20 is configured to recognize a predetermined subject (face of person) included in a wide-angle image taken by an imaging device 10 and to correct distortion of an image (face image of the person) of a part of the subject, and to associate a part of the wide-angle image with the corrected image of the part of the subject, e.g., to display a part of the wide-angle image with the corrected image of the part of the subject on a reproduction screen.