Mesh-Based Intermediate Image Synthesis for Multi-View Cameras
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for synthesizing intermediate images using multi-view square camera structures face challenges in accurately obtaining disparity vectors, reducing occlusion regions, and efficiently generating images with large disparities, particularly due to high operation times and low accuracy in distance matching and block matching processes.
Innovation Solution
The method involves distance matching to identify overlapping areas, mesh-mapping to extract vertices and construct meshes, epipolar line-based block matching to obtain disparity vectors, and image-compensating to synthesize intermediate images by integrating compensated overlapping areas, thereby reducing occlusion regions and improving accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional distance matching and block matching methods are used to synthesize intermediate images, then the synthesis process can be completed, but the operation time is high and the accuracy of disparity vector calculation is low
Solution Approach 1:
The patent segments the image processing into distinct stages: distance matching to identify overlapping areas, mesh-mapping to extract vertices and construct meshes, and epipolar line-based block matching to obtain disparity vectors. This segmentation allows each stage to be optimized independently, improving overall accuracy while managing computation time efficiently.
Solution Approach 2:
The patent performs preliminary distance matching to identify overlapping areas and construct meshes before performing the computationally intensive block matching. By preparing the mesh structure in advance, the subsequent disparity vector calculation is more efficient and accurate, reducing total operation time.
2Reliability
If multi-view square camera structure is used to obtain vivid three-dimensional effect, then the three-dimensional effect is enhanced, but occlusion regions increase and image synthesis becomes more complex
Solution Approach 1:
The patent introduces mesh structures as an intermediary representation between the multi-view images and the final synthesized image. The mesh-mapping process creates a geometric framework that facilitates the integration of multiple views while systematically handling occlusion regions, thereby reducing synthesis complexity.
Solution Approach 2:
The patent applies different processing strategies to different regions of the image based on their characteristics. Overlapping areas are processed using epipolar line-based block matching, while occlusion regions are handled through mesh-based interpolation. This localized approach manages complexity by treating each region according to its specific requirements.
3Productivity
If conventional compression algorithms are used to transmit multi-view video, then transmission can occur, but the compression ratio is insufficient for broadcast channels
Solution Approach 1:
The patent synthesizes intermediate images that represent virtual camera positions between the actual captured views. These synthesized images serve as efficient copies that convey additional view information without requiring full transmission of all possible views, thereby improving compression ratio while maintaining image quality.
Solution Approach 2:
The patent creates a universal intermediate image synthesis framework that can generate views for any camera position within the multi-view setup. This multi-functional approach allows a single set of captured images to serve multiple transmission purposes, enhancing productivity while minimizing information loss through intelligent synthesis rather than brute-force transmission.
Data Source
AI summary
The present invention relates to an intermediate image synthesizing method using a mesh based on a multi-view square camera structure.In accordance with the present invention, an accurate disparity vector may be obtained since a shared area is searched for a predetermined time interval using a distance matching and a synthesized image is generated for each area based thereon, an occlusion region is reduced by using three reference images, the synthesized image may be easily generated even for an image having a large disparity, and a converted outline may be accurately expressed, and a high three-dimensional effect may be represented due to the image conversion through the mesh based on a vertex of the outline.