研究目的
To develop an ultra-low-latency video coding method for applications like autonomous driving and virtual reality, addressing the need for high-definition video transfer with minimal delay and bandwidth usage.
研究成果
The proposed video coding method achieves ultra-low latency (0.44μs for Full-HD) with significant data compression (down to 20% of original size) and high image quality (PSNR up to 45.4dB), making it suitable for autonomous vehicles and VR devices. Future work could explore optimizations for higher resolutions or reduced complexity.
研究不足
The line-based processing cannot utilize vertical relationships in images, potentially limiting compression ratio compared to methods using 2D transforms. The optional 1D-DCT mode increases arithmetic complexity. The method is designed for specific color formats (YUV420) and pixel depth (8 bits/pixel), and may not generalize to other formats or higher resolutions without adaptation.
1：Experimental Design and Method Selection:
The study designed a line-based video coding method using processing units of 16 pixels by 1 line to minimize latency. It employed multiple prediction methods (vertical, horizontal, average, diagonal, and reference prediction), image-adaptive quantization, optimized entropy coding, and an optional 1D-DCT mode for higher compression.
2：Sample Selection and Data Sources:
Four Full-HD test video sequences at 60fps were used for evaluation, sourced from ITE/ARIB HDTV Test Materials.
3：List of Experimental Equipment and Materials:
A simulator was used to implement and test the proposed coding method; no specific hardware devices are mentioned.
4：Experimental Procedures and Operational Workflow:
The encoder processes video in compression blocks (CBs), performing prediction, quantization, entropy coding, and optional DCT. The decoder reverses these steps. Rate control was applied to manage bitrate.
5：Data Analysis Methods:
Compression ratio (CR), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM) were calculated to evaluate performance, with comparisons to H.264 Intra.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
