Full Depth Map Acquisition via Segmented Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for acquiring 3D video from 2D images are inefficient in producing high-quality depth maps, especially for arbitrary natural content, as they often result in incomplete or inaccurate depth information due to high computational complexity and limited topological complexity in single-view recordings.
Innovation Solution
A method that acquires partial depth maps, calculates derivatives of depth information, and extends these maps with non-relevant data to create a pixel-dense, spatially consistent full depth map, utilizing human perceptual constraints and models like Markov Random Fields to minimize energy functions, ensuring the depth map aligns with human depth perception.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If quantitative methods are used to retrieve depth values, then depth measurement precision is improved, but device complexity and computational effort increase
Solution Approach 1:
The patent segments the depth map into two types of pixels: boundary pixels (where depth discontinuities occur) and non-boundary pixels. Boundary pixels are processed using quantitative methods to extract precise depth values from depth cameras, while non-boundary pixels are filled using qualitative methods based on depth ordering and smoothness constraints. This segmentation allows the system to leverage the precision of quantitative methods where needed without applying the full computational complexity across the entire image.
Solution Approach 2:
The patent applies different processing quality levels to different regions of the depth map. In regions with depth discontinuities (boundaries), quantitative processing is applied to maintain precision. In regions without discontinuities (non-boundaries), qualitative processing is applied to reduce computational effort. This local differentiation of processing quality resolves the contradiction by matching the processing intensity to the local requirements of the image content.
2Measurement precision
If depth cameras are used to capture 3D images directly, then measurement precision is improved, but topological complexity of recorded scene is limited
Solution Approach 1:
The patent merges quantitative depth data from depth cameras with qualitative depth ordering information extracted from 2D images. The quantitative method provides precise depth values at boundaries, while the qualitative method provides depth ordering relationships throughout the image. By combining both approaches, the system achieves both high measurement precision and the ability to handle complex topological structures, as the qualitative component can represent arbitrary depth relationships without being constrained by the single-view limitation of depth cameras.
3Adaptability or versatility
If multiple cameras are used to capture scenes from different angles, then topological complexity is improved, but device complexity and computational effort increase
Solution Approach 1:
The patent extracts depth ordering information from 2D images using qualitative methods such as T-junction analysis and occlusion semantics. This extraction allows the system to infer depth relationships without requiring multiple physical cameras. By taking out the essential depth ordering information from single 2D images, the system achieves high topological complexity representation while avoiding the device complexity and computational burden of using multiple cameras for multi-angle capture.
4Device complexity
If qualitative methods are used to provide depth ordering information, then device complexity is reduced, but measurement precision deteriorates
Solution Approach 1:
The patent segments the depth map into boundary pixels and non-boundary pixels, applying quantitative processing to boundary pixels where precise depth values are critical for maintaining depth ordering, and qualitative processing to non-boundary pixels where depth ordering is sufficient. This segmentation allows the system to maintain measurement precision where needed while reducing overall processing complexity by using simpler qualitative methods in regions where they are adequate.
Data Source
AI summary
The invention relates to a method for acquiring a substantially complete depth map from a 3-D scene. Both depth values and derivates of depth values may be used to calculate a pixel dense depth map with the steps of acquiring partial depth map from said 3-D scene, acquiring derivates of depth information from said scene, and extending said partial depth map by adding non-relevant information to said partial depth map, creating a pixel dense full depth map being spatially consistent with both said partial depth map and said derivates of depth information.