The race to provide video surveillance products with H.264 compression has started. Manufacturers are adopting this standard compression for digital video recorders, network cameras and video encoders with the promise to reduce video data by up to 50 percent over MPEG-4 compression.
A 50 percent reduction in video data is a remarkable claim that has a big impact on the cost of ownership for a video system. Lower data rates translate to more video storage, lower network bandwidth utilization or higher quality video at equivalent data rates.
I wanted to find out for myself if H.264 can live up to this promise. I set out to find the relative compression efficiency of MPEG-4 versus H.264. In other words can H.264 really reduce video data rates without a reduction of video quality?
H.264 compression resulted from two different groups that came together to define one standard. The resulting standard is therefore known by several names. H.264 is the name used by the ITU-T which coordinates telecommunications standards for the International Telecommunication Union. The International Organization for Standardization (ISO) refers to the standard as MPEG-4 Part 10/Advanced Video Coding (AVC) since it is an extension to the MPEG-4 suite that has been widely adopted by many video security products. The physical security industry in the U.S. seems to have adopted the term H.264 as its primary reference to the standard.
The H.264 standard defines a number of new mathematical features to compress video more effectively than previous standards. Many of the features are computationally intensive and may or may not apply to specific applications. To provide flexibility, the standard defines seven different profiles. A profile defines a certain set of features to fit specific applications. Many surveillance video products are likely to implement the “baseline profile.” The baseline profile is intended for devices with limited computing resources and a requirement for low latency. Other profiles are intended for applications as diverse as video broadcasting, high-definition DVD (Blue Ray) or mobile telephony.
For the bake-off test, I applied both H.264 and MPEG-4 video encoders from Axis Communications to two typical video surveillance scenes. The first scene is a touring pan-tilt-zoom (PTZ) camera in a parking lot and the second scene is captured by a fixed camera in the front door lobby of a commercial office building. Both scenes were captured at 4CIF, 30 frames per second (fps). I used the NetVideo Device Manager Software tool to measure the data rates from each source. Through a tedious trial and error process I adjusted the compression levels of the two products to produce video of equivalent observable quality.
For both scenes, the H.264 product reduces the average data rate by approximately 50 percent.
I measured low latency of approximately100milliseconds (ms) for both compression methods. The latency is measured as the time required for the video to be captured, compressed,
transmitted over the network, decoded and rendered by a PC monitor. A latency of 100 ms is low and results in no observable delays that can compromise PTZ control.
I repeated the comparison test across different surveillance scenes and observed a common difference between the MPEG-4 and H.264 video compression. The typical blocky artifacts seen in highly compressed MJPEG and MPEG-4 video are significantly reduced in the H.264 compression.
As I increased the compression on both the MPEG-4 and H.264 video (thereby lowering both the data rates and video quality) I noticed that the “blocky” MPEG-4 artifacts become more noticeable. H.264 continues to “smooth” and eliminate the artifacts at the cost of reduced image detail.
H.264’s reduction of the blocky artifacts can be attributed to its features that reduce the size of the compression blocks to as low as 4x4 pixels and a “de-blocking” filter that “smoothes” the edges between adjacent blocks.
De-blocking is computationally expensive, requiring more processing power (and therefore cost) in the encoders deployed by the video devices.
The H.264 decoders also require more processing power. In the ‘bake-off’ comparison, the H.264 software decoder on the PC required about two times the CPU of the MPEG-4 decoder to render both the parking lot and the lobby scenes. For software applications that display many simultaneous camera views, this will have a significant impact on the selection of the PC hardware.
Although the H.264 data rate reductions require more processing power, I believe H.264 is a great step forward for video surveillance. The H.264 compression efficiencies can be used to increase video retention, lower storage costs or to improve video quality. I expect H.264 will become a pervasive compression standard for physical security with significant cost benefits for high resolution (megapixel) and frame rate surveillance systems.