Depth Map Generation Using 2D Structural Frames
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating depth maps are often inaccurate and resource-intensive, leading to suboptimal three-dimensional image quality and increased computational demands.
Innovation Solution
An apparatus and method that process an image to determine a two-dimensional structural frame, which is then used to generate a depth map, incorporating a trained artificial neural network to improve accuracy and efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional multi-view matching or scanning-based post-processing is used to generate depth maps, then depth information can be obtained, but the accuracy is suboptimal and computational complexity is high
Solution Approach 1:
The patent transitions from traditional 2D image processing to incorporating 3D structural information. By determining a 3D structural frame from 2D images and using it to guide depth map generation, the method adds a dimensional perspective that improves depth accuracy without requiring complex multi-view matching or iterative scanning processes
Solution Approach 2:
The patent performs preliminary determination of the 3D structural frame before generating the depth map. This pre-established structural information serves as a foundation that guides subsequent depth estimation, eliminating the need for computationally intensive iterative refinement processes
2Reliability
If iterative scanning approaches are used to update depth maps, then depth information may be improved in some scenarios, but computational resources are heavily consumed
Solution Approach 1:
The patent introduces a 3D structural frame as an intermediary between the 2D input image and the depth map. This structural frame acts as a mediator that provides geometric constraints and guidance, enabling more reliable depth estimation without requiring multiple iterative scanning passes that consume computational resources
3Loss of information
If multi-camera systems are used to capture three-dimensional scenes, then depth information can be obtained, but depth estimation problems result in non-ideal depth values and artefacts
Solution Approach 1:
The patent replaces the mechanical multi-camera capture system with a computational approach using 3D structural frame determination from 2D images. This substitution eliminates the depth estimation problems inherent in multi-view geometry while maintaining the ability to acquire depth information, thereby improving depth value accuracy and reducing artefacts
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An apparatus for generating a depth map for an image of a scene comprises a receiver (101) arranged to receive one or more images. A structural frame determiner (103) processes the first image to determine a two dimensional structural frame for an object in an image with the two dimensional structural frame being defined by a set of points and interconnections between the points. The two dimensional structural frame represents a projection to an image space/plane of the image of a three dimensional structural frame of the object in the scene. The structural frame(s) may for example represent a skeleton of a human or animal. A depth determiner (105) arranged to generate the depth map in dependence on the two dimensional structural frame. The structural frame determiner (103) may in many cases employ an artificial neural network to generate the two dimensional structural frame.