Making Megapixel Cameras Shine in Low Light
a&s International | Date:
The migration to IP surveillance is driven by megapixel cameras. However, low-light performance remains an unknown, as smaller and less light-sensitive pixels are crowded onto an image sensor. a&s finds out whether megapixel cameras are ready for nighttime surveillance, compared to analog models, and how components affect performance.
Megapixel is hot, spurring the migration to IP and upping demand for storage. By 2015, megapixel security cameras will account for more than 70 percent of network security cameras shipped, said IMS Research in a prepared statement.
However, megapixel is not so hot when the lights go out. Light sensitivity is challenging with more pixels crammed onto the same piece of silicon, requiring specialized techniques such as adjusting pixel density or binning. SD analog cameras are the standard to beat for low-light imaging, but more megapixel cameras are closing the performance gap.
All cameras require some light to operate, except for thermal imaging ones. Depending on the environment, it is difficult to ensure an even amount of light throughout the whole scene. “There are some analog low-light cameras that can see using starlight,” said Bob Mesnik, President of Kintronics, a US distributor.
“Well, what happens when it's a cloudy night? My point is you can't guarantee the light level in all conditions, so in the real world, we always recommend different solutions depending on the circumstances.”
Some customers need megapixel resolution for dim areas, which may require additional lighting for best results. While low-light surveillance is mature, it is not always needed. Specific applications include areas where the cost of extra lighting is prohibitively expensive or difficult to install, requiring the use of low-light cameras. Not all places with video surveillance require 24-hour footage at high quality. The average camera is good enough for most applications, limiting demand for low-light cameras.
IP Nighttime Readiness
For installations that require low-light monitoring, several things distinguish great megapixel cameras. “It's more than just the image sensor,” Mesnik said. “Besides the high-resolution sensor, you also require a good lens, good electronics that amplify the video signal and reduce noise at low-light levels, and reliability. You also require a good compression engine that provides less loss at a high frame rate.”
There is no question that network cameras look terrific in the day. “During the day, when sunlight provides a strong flux of photons, even very small pixels are enough to create great images,” said Omer Yanai, Site Manager, Vumii Imaging. “But, at night, when there is very little light, the small-sized pixels of the megapixel cameras just can't do it.” Analog nighttime imaging trumps IP. A single day/night SD camera delivers ably any time of day, simply by switching an IR cut-filter on or off. There is no magic switch for megapixel cameras at night.
Megapixel cameras can handle the night, but only under several conditions. Recording at full frame rate is possible, but the downside is more noise, particularly for CMOS sensors. “Shutter speed has more of an impact on image quality, as a slow shutter increases the amount of light,” said Alf Chang, a&s consultant and a former installer. “However, this creates ghosting and motion artifacts, so it depends on whether the user can accept a noisier image for full frame rate recording.”
IP encoding will affect network camera performance for video, which could be crucial footage at night. “Digitalizing and compressing an image cause loss of details, especially in a dynamic scene where there's a lot of movement,” Yanai said. “When dealing with a static environment and doing a detection task, using a moderate frame rate of less than 15 fps, H.264 compression will provide a very efficient video compression. But, when dealing with an identification task, from a long distance in a dynamic environment, analog video is still necessary for accomplishing the job.”
Camera performance depends a great deal on components. However, in the IP video world,sensor performance does not determine everything, as powerful processors can handle more complex algorithms.
Generally speaking, there is no difference between network and analog cameras for camera sensitivity. “As the camera's resolution increases, its sensitivity decreases inversely,” said Alex Iida, PM of Security Solutions, APAC Professional Solutions, Sony Electronics. “The market trend is seeing an increase in demand for network HD cameras.”
Whether a camera is analog or IP may not necessarily affect low-light performance. “Image sensor selection makes a bigger difference,” Chang said. “For megapixel imaging, CMOS offers more pixels but is less sensitive at 50 decibels (dB). CCD sensors are more sensitive and have better noise levels at 70 dB, but only go up to 2 megapixels or 1,080p. A 2-megapixel CCD needs a larger-format chip, more power and generates more heat, which is not practical if IR LEDs are in the mix. IR megapixel imaging is doable, but requires vents or a unique housing design for heat dissipation.”
However, the type of sensor will affect noise filtering. CMOS sensors have worse low-light performance than CCDs, but are cheaper. “If you want high performance, look for a CCD with some sort of processing smarts that adds noise filtering,” said Michael Brown , MD of VideoControlRoom, an Australian system integrator. “This gives the greatest nighttime performance, though expect a price premium.”
The most common sensor formats for high-resolution cameras are 1/2-inch and 1/3-inch sensors. “CCD and CMOS each have unique strengths and weaknesses, providing advantages in different applications. One-half inch CMOS sensors are used in many high-resolution cameras, because they provide better low-light capability, have good dynamic range and better signalto- noise ratios than CCD sensors,” Mesnik said.
The color temperature for CCDs is more accurate, while CMOS reacts more slowly to light and does not capture the best color temperature, Chang said. The differences in processing are negated with auxiliary IR illumination, as the noise from color is taken out for black-and-white images.