Seeing is Believing
Since it is difficult to determine image quality and performance from a glance at the specification sheets, side-by-side comparisons are necessary to reach a final verdict. There is no guarantee that a high-profile compressor will be better than a main-profile competitor, Chen said. “The best way to test may be using your eyes. The academic circles talk about subjective comparison and objective comparison, but really the most accurate and practical way to test real-life performance is to see for yourself. It is necessary to test both local playback and remote playback.”
“There is no substitute for the human eye,” Oliver agreed. “When testing competing products, a small library of challenging video clips should be used to assess quality at a given bit rate (or bit rate at a perceived level of video quality). The results can vary widely.”
“There are several methods for assessing quality, with most technical papers focusing on peak signal-tonoise ratio (PSNR). PSNR is easy to measure, but does not really translate to perceptual visual quality, Golston said. “Differential mean opinion score is another more useful metric that incorporates perceptual quality, but real-time measurement typically uses hardware test equipment with some upfront cost.”
A simple test is to record video on two DVRs and evaluate encoded video side by side at the same bit rate, Christensson said. “The encoded video should be a mix of standard test streams and video surveillance scenes. It can also be revealing to use ‘torture streams' to make the differences stand out more clearly. This test can include using inputs from cameras that are constantly panning and/or scenes with a lot of motion, all tested at low bit rate, such as 1.5 megabits per second (Mbps) for HD at 1,080p30.”
The actual Mbps is an indicator of how well the video has been compressed at a certain quality level, Christensson said. “Frame rate jitter is also an important measuring point. The video should be stable at the selected frame rate, regardless of video input. Some DVRs are known to drop frames when the input video becomes complex.” “The difference between a good and bad encoder is bigger if the scene is more complex and the amount of motion is larger. In a low-complexity HD scene, it may be two times; for high complexity, it may be five times or even higher,” Christensson continued
Beyond the Box
“Beyond this place of wrath and tears, looms but the Horror of the shade,” wrote 18th-century poet William Ernest Henley. The shade today for video surveillance is constantly the constraints in transmission and storage.
Storage remains a dominant factor in the operating cost of a surveillance installation, Oliver said. “Although data storage is becoming cheaper and processing power required to perform ever more efficient compression is becoming increasingly abundant, the proliferation of cameras of higher definition is acting to negate some of the benefits.“ “We can liken storage capacity to the transit infrastructure in urban cities,” Chen agreed. “No matter how many roads we manage to build, one thing that is certain is our ability to fill them up with cars.”
|
.jpg)
|
|
Andrew Yu
|
While the average cost of storage continues to drop, it is still better to reduce the space needed to store the recorded data, echoed Andrew Yu, PM of Surveillance Business Division at QNAP Systems. “By utilizing more efficient compression schemes, bandwidth and storage needed to transmit and store the video are significantly reduced.”
The bar is constantly raised and so there is always a need for better video compression, Golston said. “Moving to HD creates a huge demand for more storage, not to mention the constant need for more channels and longer retaining times. Fully exploring the capabilities of H.264 high profile using advanced algorithms, it is possible to store within the same space up to three times the amount of content at the same quality.”
The same can be applied to bandwidth, Golston added. “While network bandwidth increases over time, there are always factors driving demands at even faster rates, including dramatic increases in the number of deployed cameras, each using increasingly higher resolution.”
Latency
Another factor to consider when choosing the appropriate compression scheme is latency. Latency is the time it takes for the system to take an instruction, process it and return the result. It takes time for the encoder to compress and decompress the image, not to mention pushing it through a pipeline. It is inevitable that delays occur at various intervals, but it should be as little as possible and optimally unnoticeable to the user.
Imagine having a conversation over the phone when everything spoken is lagged for one or two seconds. The experience would be excruciating and the conversation difficult to follow, not to mention unusable for any practical purpose. This can easily be translated to a PTZ dome setting, where low latency is crucial. Long latencies make it difficult to operate PTZ equipment, Oliver said. “Much of the work in adapting broadcast codecs for the surveillance industry focuses on reducing the latency inherent in broadcast systems without reducing the quality of the video.”
H.264 provides tools within the baseline profile that ensure low latency. However, a common misconception is that while baseline is the only profile that provides tools for low-latency requirements, main and high profiles can achieve the same effect by disabling certain functions.