Mathematically speaking, reducing the number of images transmitted per second by video surveillance cameras (the frame rate) should result in a proportionate reduction in bandwidth and storage requirements.
However, those with the right technical knowledge of how video streams are constructed understand that for most video surveillance systems, a reduction in frame rate will likely only have a minor impact on bandwidth and storage usage.
Nearly all common CODECs used in video surveillance exploit the fact that, much of the time, a large proportion of any image captured is the same as the image preceding it. MPEG 4, H.264 and H.265 CODECS all leverage this fact to reduce the amount of data stored while still giving an accurate record of what occurred.
This is achieved by capturing a sequence or Group of Pictures (GOP) starting with a complete image, termed an I-Frame (Intra frame), that comprises a large amount of data because it’s a full image.
But the data captured from the next image needs only to include changes relative to the full I-Frame. This is termed the P-Frame and will usually be very small compared to the complete I-Frame, particularly when there is very little movement and change in the scene.
Because P-Frames are generally very much smaller that the I-Frames, changing the frame rate will only make a very small difference to bandwidth and storage consumption
The next image will be captured as a P-Frame and so on, until it is time to capture another complete I-Frame. It is common practice to capture an I-Frame at least once per second, so a 25fps recording would have a Group of Pictures (GOP) comprising one I-Frame then 24 P-Frames. A 12fps recording would normally consist of a GOP of an I-Frame followed by 11 P-Frames.
Hence, because P-Frames are generally very much smaller that the I-Frames, changing the frame rate will only make a very small difference to bandwidth and storage consumption. This is because it’s the I-frames that contain a fresh image of the entire scene and therefore make up the bulk of data in the video stream. There is one I-Frame per second being stored; it’s only the number of P-Frames reduced reduced as the frame-rate is reduced.
In busy environments where there is constant movement, such as train stations and town centres, the size of the P-frames may well approach that of the I-frame associated with the same moment in time. It’s therefore true to say that halving the rate at which images, made up of one I-frame and 24 P-frames, are transmitted, would have a significant impact on bandwidth and storage requirements.
Here are some screenshots showing actual data rates coming from a camera at different frame-rates in a Moderately ‘busy’ scene – ie with average amounts of detail and movement.
At 6fps the data rate is approximately 275KBps = 2.2Mbps
At 12fps the data rate is approximately 350KBps = 2.8Mbps
At 25fps the data rate is approximately 500KBps = 4Mbps
The table below shows the impact of reducing frame rates in different scenarios. It gives the typical bitrates expected from a 1080p camera viewing different scenes at different framerates.
As the data shows, reducing the frame rate by half or a quarter does not reduce the actual bitrate from the camera by the same amount. Looking at a moderately busy scene, at 25fps the bitrate is approximately 4Mbps and so it would be reasonable to expect that at 12fps, i.e. half the framerate, the bitrate would fall to 2Mbps. In fact, it would only reduce to approximately 2.8Mbps. The effect would be even greater in quiet scenes (as a lot of CCTV scenes are for long periods).
So how can you achieve maximum value from your recording solution and reduce the total cost of ownership of a video surveillance system with a large number of high definition cameras? The solution is simpler than you might think and negates the risk of some important detail of an incident not being recorded if frame rates are reduced.
It involves reducing the quality level at which images are captured. Making just a subtle change from a camera’s default image quality level will have a major impact on image files sizes – and yet operators are unlikely to notice any difference in how the images look.
Most cameras and recording devices will allow manual adjustment of ‘quality’. However, video management software (VMS) developers such as Wavestore make this possible without operator involvement.
A dynamic recording feature monitors the volume of data being recorded and estimates how long it will take to fill available recording capacity at the set frame rate. The camera’s quality level is then automatically adjusted to reduce the data rate coming into Wavestore’s VMS to ensure video is recorded and stored over the required time period – eg 31 days.