Digital Ink Occlusion Detection via Bounding Region Modeling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for rendering digital ink in videos fail to account for occluding objects, resulting in ink being rendered above foreground objects, and are computationally complex, making them unsuitable for devices with lower processing power.
Innovation Solution
An image processing apparatus that computes a model describing the pixels of a bounding region of digital ink, compares this model with subsequent frames to detect occlusion, and updates the ink accordingly, using statistical sub-models and object tracking to ensure accurate depth rendering.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a three-dimensional (3D) model of the whole scene is constructed to calculate depth and render ink at correct depth, then the ink rendering accuracy is improved, but the computational complexity increases significantly
Solution Approach 1:
The patent segments the scene into distinct depth layers by identifying occluding objects separately from the background. Instead of constructing a complete 3D model of the entire scene, the method divides the problem into: (1) detecting occluding objects in foreground frames, (2) determining their depth relative to the ink, and (3) rendering ink only in non-occluded regions. This segmentation approach achieves accurate depth rendering while avoiding the computational burden of full 3D scene reconstruction.
Solution Approach 2:
The patent extracts only the necessary information for depth rendering - specifically, the presence and position of occluding objects - rather than processing the entire scene geometry. By extracting occlusion information from video frames and using it to modify ink rendering, the system achieves depth accuracy without the computational overhead of complete 3D modeling.
2Productivity
If occluding objects are not taken into account during ink rendering, then the rendering process is simpler and faster, but the ink is rendered above occluding objects resulting in unnatural appearance
Solution Approach 1:
The patent performs preliminary detection of occluding objects in foreground frames before rendering the ink. By identifying occlusions in advance and preparing depth information, the system can render ink with proper depth relationships without adding significant computational cost during the rendering phase itself. This preliminary action ensures visual naturalness while maintaining rendering efficiency.
Solution Approach 2:
The patent introduces an intermediary occlusion detection mechanism that operates between video frame acquisition and ink rendering. This intermediary step identifies occluding objects and generates depth masks that guide the rendering process, ensuring that ink is rendered only in appropriate regions. This mediator layer achieves both natural visual appearance and computational efficiency by avoiding direct complex 3D scene analysis.
Data Source
AI summary
An image processing apparatus is described comprising a processor configured to receive a video and digital ink annotated on the video. For at least a first frame of the video, the processor is configured to compute a model describing pixels of a bounding region of the ink. For a frame of the video, the processor is configured to compute a second region corresponding to the bounding region. The processor is configured to compute a comparison between the second region and the model and update the ink using the comparison.