Stereoscopic Terrain Mesh Correction for Depth Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for converting 2D background terrain scenes to 3D stereoscopic images face challenges in accurately capturing inter-object and inner-object depth, leading to visual fatigue due to geometric errors and the 'cardboard effect', especially in large terrain scenes where inner-object depth is crucial.
Innovation Solution
A stereoscopic image generation method that involves initial mesh creation using terrain geometry, geometry error correction through projection maps and vector maps, and stereo conversion to generate accurate 3D geometry models, incorporating radial basis functions, scale-invariant feature transforms, and ray tracing for error correction and depth adjustment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If geometry based approach is used for 2D to 3D conversion, then time coherence is enforced and accurate inter-object depth data is obtained, but inner-object depth information is lost resulting in cardboard effect
Solution Approach 1:
The patent segments depth information into two distinct types: inter-object depth and inner-object depth. The geometry-based approach handles inter-object depth through mesh structure, while image-based techniques recover inner-object depth through texture analysis. This segmentation allows each method to optimize for its specific strength without compromising the other.
Solution Approach 2:
The patent creates a composite depth recovery system that combines geometry-based mesh reconstruction with image-based depth mapping. The corrected mesh provides accurate inter-object depth while texture-based depth recovery adds inner-object depth variation, creating a composite solution that leverages the advantages of both approaches.
2Measurement precision
If manual rotoscoping is used to recover inter-object depth, then accurate boundary depth is achieved, but the process is time-consuming and impractical for large terrain scenes
Solution Approach 1:
The patent replaces the manual mechanical process of rotoscoping with an automated computer vision system. The geometry-based approach automatically identifies object boundaries and reconstructs depth information through mesh processing, eliminating the need for frame-by-frame manual animation while maintaining accurate inter-object depth separation.
Solution Approach 2:
The patent changes the fundamental parameters of the depth recovery process by transitioning from manual keyframe annotation to automated continuous frame processing. By using geometric constraints and image sequence analysis, the system recovers inter-object depth for all frames simultaneously rather than requiring manual input for each frame.
3Ease of manufacture
If smooth surface geometry is prioritized for compositing, then virtual objects can be placed easily, but smaller geometric errors persist causing visual fatigue
Solution Approach 1:
The patent applies different quality standards to different regions of the terrain mesh. Smooth surface approximation is applied to large open areas where compositing is needed, while higher precision geometric reconstruction is applied to regions with distinctive features or boundaries. This local differentiation maintains both compositing ease and visual accuracy where needed.
Solution Approach 2:
The patent incorporates error detection and correction feedback loops that identify geometric inaccuracies in the reconstructed mesh. The system compares the generated mesh against the original image sequences and automatically refines the geometry to eliminate visual artifacts while preserving the overall smooth surface structure needed for compositing.
Data Source
AI summary
Disclosed herein are a stereoscopic image generation method of background terrain scenes, a system using the same, and a recording medium for the same. The stereoscopic image generation method of background terrain scenes includes an initial mesh creation step of creating an initial mesh using terrain geometry based on image sequences, a geometry error correction step of generating a projection map, detecting error regions of the initial mesh using the generated projection map, generating a vector map of the detected error regions, and generating a corrected mesh, error of which is corrected, and a stereo conversion step of generating a stereoscopic image using the corrected mesh. Since the stereoscopic image is generated based on the mesh, the mesh fits the terrain shape even though the geometry is complex. Further, time coherence can be enforced, the mesh can be edited easily, and new elements can be unseamingly composed into the terrain. Thus, it is possible to prevent a viewer who views the stereoscopic image from becoming tired.