Opportunity
The exponential growth of video content, exemplified by platforms like YouTube where over 500 hours of video are uploaded every minute, has intensified the demand for efficient video compression to reduce storage costs and conserve bandwidth. The Versatile Video Coding (VVC) standard, while offering nearly 50% better compression efficiency than its predecessor High Efficiency Video Coding (HEVC), introduces significant computational complexity due to advanced tools like multitype tree partitioning, multiple transform sizes (including non-square blocks up to 64×64), and trellis-coded quantization. This complexity, particularly in the transform and quantization modules, drastically increases encoding time, making real-time applications challenging. Existing all-zero block (AZB) detection methods from earlier standards like H.264/AVC and H.265/HEVC are inadequate for VVC because they cannot handle its unique features, such as diverse block shapes, larger transforms, and dependent quantization. Consequently, there is a pressing need for a tailored AZB detection method that can reduce encoding complexity without compromising video quality, enabling broader adoption of VVC in practical, time-sensitive scenarios.
Technology
This patent presents a novel AZB detection method specifically designed for the VVC standard. It integrates genuine all-zero block (GAZB) and pseudo all-zero block (PAZB) detection in a multi-stage framework to skip unnecessary computations. First, spatial domain GAZB detection uses a derived sum of absolute differences (SAD) upper threshold, calculated based on block size, quantization parameter, and transform type, to identify blocks that will become all-zero after quantization. Residual distributions for both square and non-square blocks are modeled using Laplacian distributions to establish this threshold. If no spatial GAZB is found, frequency domain GAZB detection applies a coefficient-level upper bound to locate the last significant coefficient in a transform block; if all coefficients are below this bound, the block is identified as GAZB, skipping quantization. For remaining blocks, PAZB detection is employed, targeting blocks that become all-zero due to trellis-coded quantization’s rate-distortion optimization. This involves estimating rate-distortion costs, comparing distortion differences and rate differences, and using a linear model to predict bit rates. The method also includes a final check on the ratio of large coefficients to avoid false positives. Pre-calculated thresholds are stored in lookup tables for efficiency, and the process reduces stages in trellis graph construction, significantly cutting transform and quantization time.
Advantages
- Achieves significant encoding time savings: average reductions of 3.83% in total encoding time under random access configuration, with up to 5% savings in transform and quantization times.
- Maintains negligible rate-distortion performance loss: average BD-rate increase of only 0.3%, ensuring video quality is preserved.
- Specifically tailored for VVC features: accommodates diverse block sizes (up to 64×64), non-square blocks, and trellis-coded quantization, unlike prior methods.
- High detection accuracy: low false negative and false positive ratios across various block sizes and quantization parameters, enhancing reliability.
- Computational efficiency: uses pre-calculated thresholds stored in lookup tables, minimizing runtime overhead.
- Scalable for real-time applications: enables faster encoding suitable for 4K streaming and other high-demand scenarios.
- Foundation for future standards: provides a model adaptable to emerging video coding technologies beyond VVC.
Applications
- Video encoding and compression software for platforms like streaming services (e.g., YouTube, Netflix) to reduce server processing loads.
- Real-time video communication tools (e.g., video conferencing, live broadcasting) requiring low-latency encoding.
- Consumer electronics such as 4K/8K televisions, smartphones, and cameras with built-in video recording capabilities.
- Video surveillance systems that need efficient storage and transmission of high-resolution footage.
- Cloud-based video processing services where computational resources are optimized for cost savings.
- Development of next-generation video codecs, leveraging the method’s principles for complexity reduction.
- Educational and research institutions focusing on video compression algorithms and performance optimization.
