What is the Impact of H.264 on NVRs and User's PCs?By: John Honovich, Published on Jun 25, 2008
The biggest concern in using H.264 lies on the impact that H.264 video streams will have on NVR servers and client PCs. If H.264 video significantly reduces the number of cameras an NVR can support or undermines the ability of operators to view video, such losses could eliminate the overall storage and bandwidth impact H.264 promises. Please see a review of potential H.264 risks for an overview.
Today I would to share 2 pieces of information:
- An explanation from Arecont Vision on why their H.264 implementation should not generate such problems
- An announcement of an independent testing program for H.264 cameras that I will run to validate performance.
Arecont Vision's Explanation of H.264 Decoding Impact
Dr. Michael Korkin, Director of Engineering at Arecont Vision shared the following explanation:
"Arecont Vision's main goal was to minimize the cost of the H.264 encoder implementation on camera as much as possible while providing a fully compliant H.264 stream, and to also reduce the corresponding computational load on the H.264 decoder, typically implemented as software running on a client computer.
The H.264 standard is comprised of multiple computational tasks, each with different computational load. The standard leaves it up to the encoder implementation to pick and choose which options are used and which are not, while maintaining full compliance with the standard. The trade-off is the stream size at a given video quality vs. the computational load in the encoder and/or the decoder. Some options contribute 1-3% to reducing the stream size while requiring an extra 50% computational load. Other options reduce the stream significantly, while requiring very little additional resources. In the encoder the extra load translates into a more expensive hardware implementation, in the decoder it translates into a higher computational load on the CPU. Moreover, within each of these options there are multiple sub-options that could vary the encoder cost and the corresponding decoder load dramatically.
We have carefully selected a subset of H.264 options and features that are important for security and surveillance applications, and then optimized their implementation on top of that, so that the cost of the encoder is very low, while the corresponding CPU load of the H.264 decoder is no higher than the MJPEG decoder load. In our own AV100 software [link no longer available] we use an off-the-shelf software decoder known as FFMPEG which produces a CPU load slightly lower than when decoding an MJPEG stream from the same camera at the same image resolution, same frame rate and same video quality. In addition, our new H.264 camera models stream at full frame rate and full resolution into QuickTime and VLC media players running on regular off-the-shelf computers."