Low Bandwidth Video Transmission via Active Appearance Model Parameter Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional video-conferencing technologies face challenges with high bandwidth requirements and lack of co-presence, leading to poor video quality and discomfort due to the absence of three-dimensional experiences, which are often achieved with expensive equipment like stereoscopic displays and multi-camera arrays.
Innovation Solution
A real-time video-conferencing system utilizing Active Appearance Models (AAMs) to track and synthesize facial features, allowing for low-bandwidth transmission of motion parallax data, enabling a three-dimensional experience with a single commodity camera, thus reducing bandwidth needs and eliminating the need for expensive equipment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If high frame-rate video transmission is used to improve video quality and co-presence, then bandwidth requirements increase significantly
Solution Approach 1:
The patent extracts only the essential motion information (displacement vectors) from full video frames for transmission. Instead of transmitting complete high-resolution frames at high frame rates, the system extracts and transmits only the differential motion data between frames, which when applied at the receiver reconstructs the high frame-rate video experience with minimal bandwidth consumption.
Solution Approach 2:
The system changes the representation parameters of video data from full-frame pixel values to compact motion displacement vectors. By transforming the data representation from spatial domain (full frames) to motion domain (displacement vectors), the patent achieves efficient compression that maintains high frame-rate capability while reducing bandwidth requirements.
2Adaptability or versatility
If binocular disparity technologies like stereoscopic displays are used to create three-dimensional experience, then equipment cost increases
Solution Approach 1:
The patent creates a virtual three-dimensional effect by copying and transforming the single-camera video feed through computational methods. Instead of using expensive multi-camera arrays or stereoscopic displays, the system generates synthetic depth cues by processing and transforming the monocular video input, providing a cost-effective alternative that achieves similar immersive effects.
Solution Approach 2:
The patent replaces mechanical/optical three-dimensional systems (multi-camera arrays, stereoscopic displays) with computational image processing methods. By substituting physical hardware complexity with software-based motion parallax and depth mapping algorithms, the system achieves three-dimensional experience using standard single-camera equipment.
3Adaptability or versatility
If motion-tracking technologies like multi-camera arrays are used to achieve motion parallax, then system complexity and cost increase
Solution Approach 1:
The patent generates motion parallax effects by computationally copying and transforming the single-camera video feed. Instead of requiring multiple physical cameras to capture different viewpoints, the system creates synthetic viewpoint changes through image processing algorithms that simulate multi-camera perspectives from a single camera input.
Solution Approach 2:
The patent substitutes mechanical motion-tracking systems (multi-camera arrays, optical tracking) with computational image analysis methods. By using algorithms to extract motion information and generate parallax effects from single-camera video, the system eliminates the need for complex motion-capture equipment while achieving similar motion parallax effects.
Data Source
AI summary
An image transmission method (and related system) for obtaining data of a local subject and processing the data of the local subject to fit a local model of at least a region of the local subject and extract parameters of the local model to capture features of the region of the local subject. The method (and related system) may also include obtaining data of at least one remote subject and processing the data of the remote subject to fit at least one of at least one region of the remote subject and extract parameters the remote model to capture features of the region of the remote subject. The method (and related system) may also include transmitting the extracted parameters of the local region to a remote processor and reconstructing the local image based on the extracted parameters of the local region and the extracted parameters of the remote region.