HDTV解码中的无损存储压缩及高效帧存结构(英文)

Lossless Memory Compression and Efficient Frame Storage Architecture for HDTV Video Decoder

提出一种用于视频解码器的无损帧重压缩方法以及参考帧存储的高效地址映射方式,可以减少解码器核心部分对外部存储器访问的读写以及行激活操作次数,并将提出的方法集成到采用MPEG-2标准的高清视频解码器中进行验证。与传统的解码器相比,在视频质量无损失且运动补偿部分读取的数据量不增加的前提下,采用本方法的解码器写入片外存储器的数据量可以减少50%左右。同时,采用将相邻的10个宏块的亮度和色度数据都存储在存储空间同一行的地址映射方法,视频解码过程中对片外存储器访问需要的逻辑换行次数可降低到传统线性存储方法的近5%。

In this work, two schemes are proposed to improve the memory performance in video decoder systems. A novel lossless frame recompression method and an efficient frame storage mapping strategy are employed to reduce the Read/Write and Row Activation operations of the external memory during video decoding processes. Verified with MPEG-2 based HDTV video decoder, the number of bytes writing to the memory is reduced by about 50% in comparison with the conventional decoder using the embedded frame recompression algorithm, without video quality degradation. Moreover, the reading bytes from off-chip memory for motion compensation are not increased. Additionally, by storing the luma and chroma components of one macroblock in one row of different banks, as well as the ten neighboring macroblocks, the logic Row Activation number can be cut down to nearly 5% of the conventional linear storage scheme.