Surface Normal Estimation Using Single-Image Points
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Three-dimensional surface reconstruction methods based on multiple images and depth maps often suffer from inaccuracies due to misalignment and low luminance conditions, leading to irregularities in the reconstructed surface.
Innovation Solution
The method estimates surface normals using only surface points derived from a single image, improving accuracy by avoiding reliance on misaligned points from multiple viewpoints and reducing the impact of low lighting conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If surface normals are estimated using surface points from multiple images, then more data is available for reconstruction, but alignment errors between images reduce estimation accuracy
Solution Approach 1:
The patent segments the surface normal estimation process by image source, computing surface normals separately for each image using only its own surface points. This avoids mixing data from misaligned images while still utilizing multiple images through subsequent processing steps.
Solution Approach 2:
The patent introduces an intermediary alignment correction step that adjusts surface points from multiple images to a common coordinate system before combining them. This mediator process corrects misalignment errors, enabling accurate combination of surface points from multiple images without suffering from the original alignment issues.
2Shape
If surface points from multiple images are used, then reconstruction completeness improves, but misalignment causes irregularities in the reconstructed surface
Solution Approach 1:
The patent performs preliminary alignment correction of surface points from multiple images before combining them for surface reconstruction. By pre-correcting the misalignment, the subsequent surface generation process receives clean, consistent data, preventing irregularities in the final reconstructed surface.
Solution Approach 2:
The patent implements a feedback mechanism where the alignment correction process iteratively adjusts surface points from multiple images based on their consistency with the final reconstructed surface. This feedback loop ensures that points from all images are properly integrated, maintaining both completeness and smoothness of the reconstruction.
3Area of stationary object
If surface points from multiple images are combined, then more surface coverage is achieved, but low luminance conditions reduce point identification reliability
Solution Approach 1:
The patent dynamically adapts the surface normal estimation strategy based on image quality metrics such as luminance. In low luminance conditions, it increases reliance on single-image surface normals; in better conditions, it more aggressively combines multi-image data, thus optimizing the balance between coverage and reliability.
Solution Approach 2:
The patent changes the weighting parameter for multi-image versus single-image surface normal contributions based on luminance conditions. When luminance is low, the weight for single-image normals increases; when luminance is high, the weight for combined multi-image normals increases, allowing optimal performance across varying lighting conditions.
Data Source
AI summary
Methods and systems for surface normal estimation are disclosed. In some aspects, a plurality of images or depth maps representing a three dimensional object from multiple viewpoints is received. Surface normals at surface points within a single image of the plurality of images are estimated based on surface points within the single image. An electronic representation of a three dimensional surface of the object is generated based on the surface normals and a point cloud comprised of surface points derived from the plurality of images.