Closing the Gap
IP low-light cameras get better and better every day — ever closer to analog cameras in terms of low-light performance.
Low-light performance (or lack thereof) has long been the knock against high definition (HD) and megapixel cameras. And for good reason because this is also one of the few areas where these cameras have been outperformed by their good, old-fashioned analog cousins.
However, this may not be the case for long — if it’s even still the case at all.
“Analog still has the upper hand for now. The technology was built around analog, and sometimes converting that performance over to IP with their megapixel technology is a little difficult,” says Todd Pinnell, product engineer for Speco Technologies in Amityville, N.Y. “IP low-light cameras are good, but not as good as standard resolution. I do not, however, think that this will be forever. IP low-light gets better and better every day.”
Over the last 12 to 18 months, technological improvements allowed HD and megapixel cameras to creep ever closer to analog cameras in terms of low-light performance.
“People used to subscribe to the notion that denser images can’t gather as much light,” says Steve Surfaro, security industry liaison for Axis Communications, based in Chelmsford, Mass. “But what was true a year and a half ago is not true anymore.”
The Perfect Storm
So just how have manufacturers been closing the gap between the low-light performance of high-resolution cameras and that of analog cameras?
One of the biggest technological advancements has involved a culture shift, as camera manufacturers have moved away from simply slowing down the shutter speed to let in more light, while moving toward “actually improving the technology,” Pinnell says.
To fully understand how far HD and megapixel cameras have come, it’s important to understand what’s been holding their low-light performance back for so long. And to get to the heart of that, you have to get into the “guts” of the camera, specifically three parts: the lens, the sensor and the encoder. Advancements in all three have combined to form a perfect storm that has allowed the relatively rapid improvements in low-light cameras in recent months.
“Several technological breakthroughs in megapixel imaging have come together at the right time,” says Steve Carney, senior product manager of IP cameras and encoders, Tyco Security Products, based in Westford, Mass. “The key was that the sensitivity of CMOS sensors had to be improved.”
Historically, HD and megapixel cameras have used CMOS sensors, which are better at processing large amounts of pixels but have always delivered disappointing low-light performance. Analog cameras, on the other hand, have relied on CCD sensors, which are far more efficient in low-light situations. The drawback is that CCD sensors are limited in terms of the number of pixels they can handle. This disparity meant end users, integrators and installers had to choose between high-resolution images and low-light capability.
However, CMOS sensors have become widely used in the consumer market, which has driven the price down (or vice versa). Consumer demand has also driven improvements in CMOS technologies, which has coincided with improvements in processors and lenses.
“Faster” lenses now allow more light into the camera without slowing down shutter speeds. IR-corrected lenses continue to improve.
The processor is the glue that holds imaging together, Surfaro says. After all, if the camera isn’t strong enough to process the signal and manipulate the image (to account for a defective pixel or remove noise from the scene, for example), then what comes across the network to the headend is going to be nothing short of disappointing.
“Performance doesn’t come easy,” he says. “Imager, processing, network performance and the ability to sustain and maintain images — it all takes a little work. Do any one of them poorly, and the whole process is affected.”
The main factor that determines the performance level of a low-light camera is the confluence of the “big three” technologies, which Carney calls a “dark art.”
“There’s a lot of ‘secret sauce’ in good low-light cameras,” he says. “When the lens, sensor and processor mix together you can have very different outcomes from one lens, sensor or processor to another.”
End Users in the Driver’s Seat
As is often the case in the security industry, much of the technology improvement in low-light cameras comes directly from the consumer market. For an example of what the industry can expect as far as improvements in low-light cameras, consider the differences between point-and-shoot megapixel cameras ten years ago and today, says Willem Ryan, product marketing manager for Fairport, N.Y.-based Bosch Security Systems.
“We’re not in the early adoption stage. Security follows the consumer market, and we’ve seen the proliferation of megapixel cameras there,” he says. “With the first megapixel cameras, low-light pictures were horrible, but we’re starting to see improvements fall up to security. It’s usually a year or two behind, so in the next year or two, you’re going to see dramatic improvements.”
Another improvement in IP low-light cameras is video analytics, which is an area where IP (including HD and megapixel) leaves analog in the dust.
“IP low-light performance is catching up, and you also get the benefit of increased resolution and intelligence, which isn’t available on the analog side,” Carney says. “It is turning into a market where it is becoming just an issue of cost. If you’re focused on the least expensive option, you’ll choose analog. If you consider the quality and lifecycle, IP is the choice. Performance is no longer an inhibitor to customers making decisions.”
One of the big problems with low-light cameras in the past has been noise. H.264, which is more or less the standard for video compression, relies on taking keyframes, meaning it only sends information that’s changed within a scene back to the headend. In a clear, well-lighted scene, there’s usually very little change, if any at all. In a low-light scene, the camera’s sensor and encoder interpret noise within an image as motion. Therefore, cameras that perform poorly drive up network congestion and storage requirements, Carney says.
“Fuzz is motion,” he says. “Say you have a 1 megapixel camera recording at 720p. That may transmit video in a well-lit scene at 2 to 4 megabits a second. In a night scene with nothing moving, but with a lot of noise in the scene, it may be putting out 8 to 10 megabits a second.”
It follows that cameras that perform well in low-light situations will reduce the amount of noise in an image and by extension lower bandwidth and storage requirements. In other words, the savings on the back end could possibly offset the cost difference between an IP and an analog camera.
Determining the performance end users require from a low-light camera is where the whole sales process starts. Before rushing to any judgment, Ryan advises installers and integrators to start by using the Johnson Criteria, or DCRI (detect, classify, recognize and identify).
“Knowing which of the DCRI criteria they need will drive the camera they choose,” he says. “Do they need to recognize faces, or just that it’s a human, or just motion?”
Just as important is knowing the available light at a location at any given time during the day, Surfaro says.
“Is it a residential area where there’s not a lot of light at night, or is it a train station with an over-platform light on all night?” he says. “The available light is going to play a huge role in the ultimate decision about which cameras to deploy.”
In the world of low-light cameras, as in many areas of the security industry, there are no standards by which manufacturers determine the specs they list on their product information sheets. It stands to reason then, that two cameras with identical performance specs listed can, and often do, perform very differently in real-world situations.
The only true way to determine the best camera for any situation is to conduct a shootout between cameras from several manufacturers, Surfaro says.
“There are no standards for lux levels, so you can’t rely simply on spec sheets,” he says. “One may look better than another on paper, but the other may actually perform better because that manufacturer was conservative and set realistic expectations.”
One example of “specsmanship” is the distance from a camera at which lux levels are mentioned. A camera that performs well at close range may perform much worse as the distance between the camera and the scene it is recording increases, Ryan says.
“You can’t tell how far away from the camera that lux level was measured, so there’s no way other than a shootout to know exactly what kind of image you’ll get with the actual light level at a location,” Ryan says.
Slower shutter speeds have been a method long employed by IP camera manufacturers to increase their cameras’ low-light performance — with mixed results. The slower the speed, the greater the potential for blurring in an image, which can render video completely useless. This factor aside, Pinnell says varying shutter speeds is another trick some manufacturers can use to boost their cameras’ specs.
“The biggest issue I can see is knowing which technology is being for used for the camera. Is it a true intensifier, or is it just a cheap camera opening up the shutter a little bit longer?” Pinnell says. “Buying the correct camera and/or technology will save integrators and end users from a lot of headaches in the future.”
Understandably, a shootout may not be feasible or even possible for every situation. In that case, there are a few other ways to know you’re choosing a good low-light camera, Carney says. One of these is to look at the bandwidth calculations manufacturers report based on testing in different scene types and different lighting settings. These calculations are quoted in a way that allows installers and integrators to estimate storage requirements associated with a particular camera.
“There’s a little more truth in advertising with bandwidth calculation than simply with specs,” he says. “So if there’s no clear data presented in a well-constructed bandwidth calculation, than may be a good indicator.”
Even without performing a shootout, Ryan advises installers and integrators to familiarize themselves with a variety of manufacturers’ cameras and with low-light technology itself.
“Each manufacturer designs a camera to work out of the box, but depending on the settings, you’ll get different performance levels,” he says. “Don’t be afraid to play around with cameras to see how they can perform. Nothing can replace that experience.”
Based on the breadth and depth of technologies available today, Ryan says there’s a right camera for any application. The trick is to determine what that is — and it may not be a low-light camera.
“There are a lot of tools out there — low-light cameras, IR, white-light illumination, thermal — so there’s no excuse to not provide a great image 24x7,” he says.
Naturally, all the improvements in low-light technologies have led to greater demand for and adoption of low-light cameras from end users, whose acceptance of the technology might be surprising, says Cheryl Bard, product marketing manager for Fairport, N.Y.-based Bosch Security Systems.
“These days, a day/night camera is almost default by choice, with IR sensitive as standard,” she says.
Among the core verticals where low-light cameras have been most used are industrial facilities, utilities, critical infrastructure, transit and transportation, with large industrial and petrochemical facilities as the “power users,” according to Willem Ryan at Bosch Security Systems. However, the technology is finding its way into applications and verticals where it hasn’t traditionally been considered.
These new opportunities include data centers, city surveillance, public safety and education — both K-12 and at the college level — or a combination of any of these. One example of this “teamwork” approach, says Steve Surfaro at Axis Communications, might be public safety organizations using video from cameras placed on playgrounds that also capture part of the surrounding streets.
“Because of the advancements in the imager and codec technology, low-light or day/night cameras are being utilized in areas not currently covered by this technology,” says Edward Wassall, director of IP product and business development for Samsung Techwin of America, based in Carson, Calif. “Users can now install this technology in areas such as outdoor with megapixel technology and receive a very resolute image at very low light ratings.”
In fact, outdoor surveillance is one of the major applications driving the increased adoption of low-light cameras, says Todd Pinnell of Speco Technologies.
“Low-light cameras are used in almost every outdoor application,” he says, adding that low-light cameras are his company’s most popular cameras and “are going into almost every application that in the old days were fitted with an IR camera.”
So why are end users buying in (and manufacturers promoting low-light cameras more than in the past)? The answer, Pinnell says, is simple.
“My personal feeling is that it’s because the technology is actually good now,” he says. “Many people gave up on low light cameras after the first go-round a few years ago and went back to IR cameras. IR is still the way to go for a dead black situation, but now with the improvements being seen with low-light cameras, coupled with the ‘color by day/color by night’ abilities of those cameras, many manufacturers are pushing these as the way to go.”
|Thermal Heats Up|
One subset of low-light cameras that’s generating a lot of buzz is thermal imaging. Long considered to be solely for military and government applications, thermal cameras have made inroads to other areas, thanks in large part to lower prices.
“Thermal has come into its own and it’s getting looked at harder,” says Bosch’s Willem Ryan. “It’s always been sexy, but way too expensive. In the last two years, entry-level sensors have made thermal more affordable and reasonable.”
Thermal cameras operate by sensing heat emitted from objects. All objects, whether living or not, emit heat that these cameras can detect based on their ability to sense temperature variations down to 1/20 of a degree. Because heat is present at any time of day, thermal cameras can operate in both sunny and dark situations. Naturally, total darkness is where they really excel, says John Romanowich, chief executive officer of Princeton, N.J.-based SightLogix Inc.
“Thermal cameras have become a good choice for seeing in the dark,” he says. “At night, background objects tend to be cooler than a person. Therefore, under ideal conditions, people are well-emphasized at night because they appear brighter in the image than the background, standing out even in zero light.”
The resulting images generated by thermal cameras are rich in data, with temperature variations presented in more than 16,000 shades of gray. The challenge is that a thermal camera has to compress these thousands of shades of gray down to the 256 shades the human eye is capable of seeing. In the past, this compression process involved combining areas that were close in temperature together. Traditionally, this has meant blurry images that are seemingly lacking in detail — potentially concealing intruders.
Today’s cameras rely on more intelligent processing that emphasizes small variations between objects and background to “exaggerate” fine details, resulting in a clearer image in contrast to other image features, Romanowich says.
Another challenge with thermal cameras has been the effect of rain, fog, humidity and other environmental conditions. Often, these factors could allow an intruder to blend in to the background of a scene. Thanks to technology, this is no longer the case, Romanowich says.
“Intelligent processing can be used to remap the image to emphasize the small temperature differences in the hotter objects, presenting an image that approaches a black-and-white video quality, which is more comfortable to the eye and will better emphasize potential intruders,” he says.
With technology improvements has come greater ease of installation, says Axis Communications’ Steve Surfaro.
“There are some differences, but to deploy a thermal camera just requires a little increased application knowledge,” he says. “Once you have it installed, it’s just another network camera.”
While there are some applications where simply detecting people is the goal, such as military or border patrol, thermal imaging works well as a standalone solution, Ryan says. But because those situations aren’t very common in the civilian world, he advises using thermal cameras in conjunction with other solutions.
“Thermal isn’t the one and only, but if you use it to complement a visible or camera with or without IR or other illumination, it works well within a total solution,” Ryan says.