Backward 3D View Generation with Depth Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current 3D display methods, particularly in stereoscopic approaches, face challenges with accurate pixel projection and handling of occlusions and de-occlusions, leading to complex processing and potential artefacts due to integer depth values and approximation methods.
Innovation Solution
A method for generating image views in 3D displays that associates pixels based on a texture map and depth map, using a projecting function that considers depth values and shifting angles, and compares potential depth values to actual values to accurately assign texture values, reducing complexity and artefacts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If forward processing approach is used to generate view pixel values from input pixel and depth values, then the computation is simpler, but the precision is reduced due to integer depth values and pixel positions may not be assigned accurately
Solution Approach 1:
The patent inverts the traditional forward processing approach by implementing backward processing. Instead of projecting input pixels forward to view pixels (which causes precision loss with integer coordinates), the method starts from view pixels and back-projects to find corresponding input pixels. This inversion allows using floating-point depth values for accurate depth comparison and selection, resolving the precision issue while maintaining computational feasibility through a different algorithmic approach.
2Ease of manufacture
If forward processing approach is used, then computation is easier, but occlusions and de-occlusions must be handled through complex post processing
Solution Approach 1:
The patent applies preliminary action by handling occlusions and de-occlusions during the view generation process itself rather than through subsequent post-processing. The backward processing method inherently manages these issues by comparing depth values during pixel selection, determining the correct pixel assignments before the view is finalized. This eliminates the need for complex post-processing steps to resolve occlusion artifacts.
3Reliability
If backward processing approach is used to generate image views, then all pixels in views are visited and no holes are left, but the depth of the pixel in the input image is not known accurately leading to approximation artefacts
Solution Approach 1:
The patent implements feedback by using the depth map information to guide the backward processing. When back-projecting from a view pixel to potential input pixels, the method retrieves and compares the actual depth values from the depth map for each candidate pixel. This feedback loop allows accurate depth comparison to determine which input pixel should be assigned to the view pixel, eliminating approximation artifacts while maintaining the completeness advantage of backward processing.
Data Source
AI summary
There is described a method for generating an image view from at least one input image for a 3D display using a backward processing enabling post processing for handling holes in the image view.