Selective Pixel Rendering for Stereo Reprojection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current re-projection techniques for creating 3D images are slow and result in undesirable artifacts, such as holes and distortion, due to the need for extensive processing and the reliance on stretching or compressing pixels, which is particularly problematic for complex scenes and non-stereoscopic images.
Innovation Solution
A system and method that uses a depth map and intraocular spacing to generate a disparity map and distortion map, allowing for the creation of a new view without building geometry or using a renderer, and selectively renders only affected pixels to minimize artifacts, thereby reducing processing time and resources.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If re-projection techniques are used to create stereoscopic images from a single rendered view, then processing time is reduced, but image quality deteriorates due to artifacts such as holes and distortion
Solution Approach 1:
The patent applies local quality by differentiating between two types of pixels: re-projected pixels (from the first view) and rendered pixels (for the second view). Instead of uniformly processing all pixels, the system selectively renders only those pixels that are occluded in the first view but visible in the second view, thereby maintaining high image quality while minimizing processing time.
Solution Approach 2:
The patent segments the pixel set into two distinct groups: pixels that can be re-projected from the first view and pixels that must be rendered from the second view. This segmentation is achieved by comparing depth values and occlusion relationships between the two views, allowing the system to process only the necessary subset of pixels through rendering.
2Manufacturing precision
If all pixels are rendered to create a second view, then image quality is maintained, but processing resources are wasted
Solution Approach 1:
The patent applies partial action by rendering only the minimum necessary subset of pixels required to create the second view. Instead of rendering all pixels, the system identifies and processes only those pixels that are occluded in the first view, performing exactly the amount of rendering needed without excess.
Solution Approach 2:
The patent extracts and processes only the specific subset of pixels that require rendering, separating them from the larger set of pixels that can be handled through re-projection. This extraction is based on depth comparison and occlusion analysis, allowing the system to isolate and process only the necessary elements.
3Reliability
If pixel stretching or compression is used to fill re-projected areas, then holes are avoided, but undesirable artifacts are introduced
Solution Approach 1:
The patent uses copying by re-projecting pixels from the first view to the second view when appropriate. Instead of stretching or compressing pixels, the system copies pixel data from the original rendered view and maps it to the corresponding location in the second view using depth information, thereby maintaining image quality without introducing artifacts.
Data Source
AI summary
Systems and methods are provided for minimizing re-projection artifacts. From a re-projection method, a distortion map determines areas of excessive stretching or compressing of pixels, which are then rendered. The re-projection and render are composited to create a new stereo image. Systems and methods are also provided for a “depth layout” process, whereby a new offset view of a 2D plate is created, allowing a stereo pair of images. Custom geometry that approximates the scene is created and 2D rotoscoping is employed to cut out shapes from the geometry. A depth map is created, from which an offset image may be re-projected, resulting in a stereo pair. Systems and methods are also provided for a fast re-projection technique, which creates a novel view from an existing image. The process generates a disparity map and then a distortion map, which is then used to create the new view.