Description:
Background
Modern compression systems use a set of fixed mathematical tools to remove unnecessary information from each video frame. While these tools work well, they treat most video content in the same way and cannot adapt to the unique characteristics of each scene. As video resolutions, frame rates, and streaming demands continue to increase, there is a need for compression technologies that maintain low complexity while delivering the coding efficiency of adaptive transforms.
Technology Description
USC researchers have developed a process that learns patterns within the video and adjusts its processing to better match the content being compressed. This allows the system to represent the same visual quality using fewer bits of data. Testing within the latest video coding framework showed that INT-DTT+ can reduce the bitrate required for video compression in intra frames by more than 3% compared to existing methods while maintaining similar processing complexity. This delivers the same quality using less bandwidth and storage, potentially lowering infrastructure costs and improving streaming performance.
Benefits
- >3% BD-rate improvement in intra frames relative to VVC MTS baseline, enhancing compression efficiency
- Lowers operating costs for streaming, video sharing, and cloud video platforms.
- Reduces bandwidth and storage requirements while maintaining video quality.
- Improves encoding and decoding efficiency for software and hardware implementations.
Stage of Development
- Lab prototype
- Provisional patent application filed
Applications
- Video streaming platforms
- Video surveillance and security systems
- Consumer electronics and smart TVs
Related Publication
https://arxiv.org/pdf/2511.17867