论文部分内容阅读
为了提高H.264/AVC变换和量化部分硬件实现的速度,通过分析整数离散余弦变换(DCT)和量化模块的速度优化设计,提出一种算法的并行流水线处理结构。这种结构可同时处理16个不同数据类型的像素点(亮度或色度),降低了计算复杂度,避免了解码端的失配问题。实验结果表明,优化后的算法吞吐量达到3 564Mpixel/s,PSNR只降低了约0.02dB,满足实时性的要求,获得了比以往标准更好的编码性能。
In order to improve the speed of hardware implementation of H.264 / AVC transform and quantization, an algorithm of parallel pipeline processing architecture is proposed by analyzing the optimal design of integer discrete cosine transform (DCT) and quantization module. This kind of structure can deal with the pixel point of 16 different data types at the same time (brightness or chroma), Have reduced the computational complexity, Have avoided the mismatch problem on the decoder side. Experimental results show that the optimized algorithm achieves a throughput of 3 564Mpixel / s and the PSNR only reduces by about 0.02dB, meeting the requirements of real-time performance and obtaining better coding performance than the previous standard.