Triangulation-based Adaptive Subsampling of Dense Motion Vector Fields

Simple SummaryContent extracted from patent full text and abstract with AI.

This invention presents a new method and apparatus for efficiently compressing and reconstructing dense motion vector fields in image or video coding. Instead of transmitting motion vectors for all positions in a dense grid or signaling each non-regular sample position, the method adaptively selects only key points where motion changes significantly using triangulation (such as Delaunay triangulation). New sample points and their associated motion vectors are derived iteratively, providing dense sampling where needed and sparse sampling in homogeneous areas, all without needing to transmit explicit positions for each vector. This leads to improved compression efficiency in video encoders/decoders, reducing bitstream size while maintaining (or improving) motion fidelity.

Use CasesContent extracted from patent full text and abstract with AI.

• Video compression and encoding for streaming platforms (e.g., YouTube, Netflix)
• Video conferencing systems to reduce bandwidth usage while maintaining quality
• Mobile device video recording and playback to save storage and transmit efficiently
• Video surveillance systems where efficient storage/transmission is critical
• Virtual reality and augmented reality video for efficient encoding of immersive environments
• Any application utilizing motion-compensated interpolation or optical flow (e.g., slow-motion, frame interpolation, video stabilization)
• Broadcast video transmission and reception, where bandwidth is valuable

BenefitsContent extracted from patent full text and abstract with AI.

• Reduced bitrate for transmitting videos, leading to lower storage and transmission costs
• Enables higher video quality at the same bitrate, or the same quality at a lower bitrate
• Reduces blocking and other artifacts produced by block-based motion estimation
• Adaptively samples motion fields to better preserve details in areas of complex motion
• No need to explicitly signal every motion vector position, simplifying the bitstream and reducing overhead
• Supports progressive and scalable video transmission; can adjust quality dynamically by number of iterations
• Flexible and compatible with existing video codec frameworks, allowing straightforward integration

Technical Classifications (CPCs)

Inventors & Applicants

Patent Abstract

The present disclosure relates to an apparatus and a method for providing a plurality of motion vectors related to an image coded in a bitstream, comprising obtaining a set of sample positions within the image, obtaining respective motion vectors associated with the set of sample positions, deriving an additional motion vector based on information coded in the bitstream, determining an additional sample position located within a triangle, which is formed by three sample positions of the set of sample positions, based on the respective motion vectors associated with the three sample positions, the triangle not including any other sample position of the set, adding the additional sample position to the set of sample positions, and associating the derived additional motion vector with the additional sample position. Such method and apparatus provide for an efficient motion flow compression and reconstruction enabling to recover some sample positions without signaling them in the bitstream.