Optical Flow Boundary Pixel Hint Expansion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Pyramid optical flow processing faces challenges near object boundaries, where erroneous hints can lead to inefficient computations and incorrect optical flow mapping, especially when objects in the foreground and background move at different velocities.
Innovation Solution
The method expands the source of motion vector hints by incorporating neighbor superpixels, allowing boundary pixels to receive hints not only from directly associated superpixels but also from neighboring ones, thereby capturing a broader pool of candidate areas for accurate pixel matching.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If motion vector hints are obtained only from directly associated superpixels in pyramid optical flow processing, then computational complexity is reduced, but accuracy deteriorates near object boundaries where foreground and background objects move at different velocities
Solution Approach 1:
The patent segments the hint source by distinguishing between directly associated superpixels and neighboring superpixels. Boundary pixels receive hints from multiple segmented sources (direct superpixels and neighboring superpixels), while non-boundary pixels use only direct superpixels. This segmentation allows the system to increase accuracy specifically where needed (at boundaries) without applying the computationally expensive operation globally.
Solution Approach 2:
The patent applies local quality by making the hint acquisition strategy adaptive to local image characteristics. The system identifies boundary pixels and applies a different hint acquisition method (using neighboring superpixels) specifically to these local regions, while using the standard method (direct superpixels only) for non-boundary regions. This ensures high accuracy at critical boundary locations without unnecessarily increasing computational load across the entire image.
2Reliability
If the system processes all pixels with the same method, then implementation simplicity is maintained, but performance deteriorates at object boundaries
Solution Approach 1:
The patent introduces dynamics by making the processing method adaptive rather than static. The system dynamically determines whether to use neighboring superpixels based on whether a pixel is identified as a boundary pixel. This dynamic adaptation allows the system to maintain simplicity for most pixels while automatically increasing complexity only where needed for boundary pixels, improving reliability without making the entire system complex.
Solution Approach 2:
The patent applies preliminary action by pre-identifying boundary pixels before the optical flow computation. By marking boundary pixels in advance, the system can then apply the appropriate hint acquisition method (using neighboring superpixels) specifically to these pre-identified pixels during the main processing loop, rather than having to determine boundary status for each pixel individually during computation.
Data Source
AI summary
Disclosed are apparatuses, systems, and techniques that may perform methods of pyramid optical flow processing with efficient identification and handling of object boundary pixels. In pyramid optical flow, motion vectors for pixels of image layers having a coarse resolution may be used as hints for identification of motion vectors for pixels of image layers having a higher resolution. Pixels that are located near apparent boundaries between foreground and background objects may receive multiple hints from lower-resolution image layers, for more accurate identification of matching pixels across different image levels of the pyramid.