3D Spherical Image Mapping for Seamless Stitching
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in generating a spherical image from multiple images captured by different cameras with overlapping fields of view, leading to issues like disparity and difficulty in stitching images, especially in applications where cameras are mounted on restricted spaces like hard hats, resulting in unnatural-looking images with artifacts.
Innovation Solution
An apparatus and method that select an image to be foreground based on the orientation and angle of view of a virtual camera, map multiple images onto a three-dimensional object, and perform perspective projection to generate a plane image with a wider angle of view, allowing for seamless stitching and display of images without alignment or blending, by determining the rendering order of images based on their capturing directions and the virtual camera's orientation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of stationary object
If multiple images are captured by cameras with overlapping fields of view to generate a spherical image, then the coverage area is improved, but image stitching difficulty increases due to disparity
Solution Approach 1:
The patent maps multiple two-dimensional images onto a three-dimensional spherical surface, using the third dimension (radial distance from sphere center) to naturally resolve overlapping regions. Each image is projected onto the sphere from its respective camera position, and the spherical surface automatically handles the spatial relationships, eliminating the need for complex 2D stitching algorithms.
Solution Approach 2:
The spherical surface acts as an intermediary medium between multiple camera images. Instead of directly stitching images together in 2D space, the patent uses the spherical surface as a common reference frame that naturally accommodates images from different viewpoints, orientations, and distances, simplifying the integration process.
2Adaptability or versatility
If cameras are mounted on restricted spaces like hard hats to capture spherical images, then the adaptability is improved, but image alignment accuracy deteriorates due to large disparities between camera positions
Solution Approach 1:
By transitioning from 2D image plane coordination to 3D spherical surface mapping, the patent naturally handles large spatial disparities between cameras mounted on restricted surfaces. The spherical coordinate system accommodates different positions and orientations without requiring precise alignment, as each image is mapped from its own viewpoint onto the common spherical surface.
Solution Approach 2:
The patent changes the coordinate system parameters from conventional 2D Cartesian coordinates to spherical coordinates (radius, latitude, longitude). This parameter transformation allows images captured from widely separated positions on restricted surfaces to be naturally integrated, as the spherical coordinate system is inherently suited for representing views from multiple positions around a central point.
3Productivity
If traditional image stitching methods are used to combine multiple images, then the productivity is improved, but image quality deteriorates due to artifacts and unnatural appearance
Solution Approach 1:
The patent avoids traditional 2D image stitching by mapping images onto a 3D spherical surface, where overlapping regions are naturally resolved through the spherical geometry. This eliminates common stitching artifacts such as ghosting, misalignment, and unnatural boundaries that occur when forcing 2D images to align, while maintaining processing efficiency through automated spherical projection.
Data Source
AI summary
An apparatus includes circuitry configured to: select, from at least two images captured in different image-capturing directions and with image-capturing ranges overlapping with each other, an image to be at foreground as viewed from a virtual camera based on an orientation or an angle of view of the virtual camera and the image-capturing directions of the at least two images; map the at least two images onto a three-dimensional object to generate a virtual image, in which the at least two images overlap with each other, having a wider angle of view than the at least two images; and perform perspective projection on the virtual image using the virtual camera, to generate a plane image, based on the selected image to be at the foreground, as a display image.