Depth Order Metadata for Multi-View Rendering Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Multi-view 3D rendering techniques face errors due to depth map uncertainties, causing objects of similar depth to overlap incorrectly, particularly in scenarios like a soccer player's foot and the ground, which requires additional data transfer and processing to correct.
Innovation Solution
Incorporating qualitative depth information to determine a depth order for objects, where objects are categorized based on their relative depth, reducing data requirements and processing by using pre-defined or extracted knowledge about object positions, such as the soccer field always being behind players, and storing this information separately from depth maps.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If depth maps are used for multi-view rendering, then depth information is obtained, but rendering errors occur due to depth uncertainty and estimation errors
Solution Approach 1:
The patent segments the scene into multiple depth orders (first depth order, second depth order, etc.) based on qualitative depth information. This segmentation allows the rendering system to handle objects at different depth levels separately, preventing rendering errors that occur when depth maps incorrectly determine the depth relationship between objects with similar depths.
Solution Approach 2:
The patent performs preliminary classification of objects into different depth orders before rendering. By determining the depth order of objects in advance based on qualitative depth information (such as object type, position, or semantic information), the system establishes a reliable depth hierarchy that prevents rendering errors during the actual rendering process.
2Reliability
If additional qualitative depth information is stored to correct depth ordering errors, then rendering accuracy improves, but data transfer and processing requirements increase
Solution Approach 1:
The patent applies qualitative depth information selectively to objects that require depth order correction. Instead of adding comprehensive depth data for all objects, the system identifies specific objects where depth ordering is uncertain or incorrect and applies qualitative depth information only to those cases, minimizing additional data requirements while maintaining rendering accuracy.
Solution Approach 2:
The patent changes the parameter representation from continuous depth values (which require high precision and generate large data) to discrete depth order categories (first depth order, second depth order, etc.). This parameter transformation reduces the data volume significantly while maintaining sufficient information for correct rendering, as depth orders only need to indicate relative depth relationships rather than precise measurements.
Data Source
AI summary
A method for storing multi-view data with depth order data. The method comprises obtaining image frames of a scene from an imaging system with a plurality of cameras, obtaining depth maps from the imaging system and/or the image frames and obtaining qualitative depth information relating to the depth of at least one object present in the scene relative to other objects in the scene, the qualitative depth information being additional to the information conveyed by the depth map. A depth order is determined for a set of at least two objects present in the scene based on the qualitative depth information, wherein the depth order determines the depth of an object relative to other objects with different depth orders. The image frames of the scene, the corresponding depth maps and the depth order for the objects in the scene are then stored as the multi-view data.