IP Cameras and Low Light

From CCTV Information
Revision as of 10:43, 4 January 2013 by Gb003162 (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

What has changed and how to choose; by Steve Carney

Whether or not it’s deserved, IP cameras have developed a reputation for poor performance in low light. And because low-light conditions exist with nearly all camera installations, this perception has inhibited some organisations from adopting the technology, in spite of its many strategic and far-reaching benefits.

About the author: Steve Carney is Director of Product Management, Tyco Security Products, responsible for American Dynamics products.

The challenges have had less to do with delivering the video over an IP network than with the CMOS sensors that most IP cameras use. Historically these sensors have been able to deliver higher megapixel resolution, but they weren’t able to match the CCD sensors often used in analogue cameras for low-light performance. Because of this, some IP cameras generate grainy images at low light, resulting not only in decreased picture clarity but also in higher bandwidth usage and increased storage, since compression techniques interpreted the graininess as motion in the scene. The past 12 months, however, have seen some advancements in technology that have enabled the IP camera to function as a more useful tool in low-light situations. Camera manufacturers have been able to leverage advances in sensors, encoding and processing power born from the automotive and other industries to resolve many of these issues. And High Profile H.264 compression helps to more effectively manage bandwidth usage, whilst more sensitive elements in the sensors provide higher-quality images.

Companies hoping to benefit from these advances should look for the following features in low-light IP cameras:

True day-night: True day-night capability is achieved by using an IR cut-filter mechanism (IRCF). Many of the sensors used in today’s security cameras are sensitive to both visible light (380nm~740nm), essentially what the human eye can see, as well as near infrared light (750nm~1100nm), which is emitted from sources such as sunlight, moonlight, halogen fixtures, etc. Unfortunately, to produce accurate colours, most of the IR light needs to be blocked or filtered out. Many cameras do this by using an IR Cut Filter, which sits in front of the sensor like a pair of ‘sunglasses’. With this True Day Night /IRCF feature the camera is equipped to remove these ‘sunglasses’ when the light levels drop below a certain threshold. This allows more of the visible light AND available IR light to get to the sensor, dramatically improving low-light performance. Because of the filtering or blocking of the IR light, colour accuracy usually suffers in this mode. Most implementations also remove all colour information, yielding a black and white image that is not only vastly more usable but also cleaner without the chroma noise. This improves image clarity whilst decreasing noise or graininess in the image, which can be interpreted as motion in the encoder. When the encoder compresses an image with a higher level of motion or noise, the bandwidth consumption can rocket.

One other true day-night technology differentiator is what manufacturers do when the IRCF is in the ‘out’ or removed position. The better ones replace the IRCF with a piece of dummy glass to help minimise the spectral offset between visible and IR light. This helps keep the image in focus and retain auto focus stability. Be aware there are cameras that have a ‘Quasi’ or soft Day Night (SDN) feature; this is not the same thing as True Day Night. These type of cameras will go black and white when the light levels are low but they have no mechanism to remove the IRCF ‘sunglasses’. This means they cannot take advantage of all the visible and IR available light. Marketing literature has made this a source of confusion in our industry; make sure you know which day-night implementation your cameras of interest have.

Also note that the use of IR corrected lenses is highly recommended with True Day Night cameras. Since visible and IR light are in different wavelength bands they focus at different points, creating a potential problem of focus shift when switching between colour and black and white modes. A good quality IR corrected lens can compensate, meaning fewer service calls to correct the focal points.

Digital Slow Shutter: Digital Slow Shutter (DSS) is a feature where the camera’s electronic shutter is slowed to below normal speed to allow additional light to be captured by the sensor. It works very well in improving low-light performance but has some side effects including image noise and a potential to blur moving objects. If a camera has this feature it is important to see how well it was implemented, as not all executions of slow shutter are equal. Besides the noise and blur, be on the lookout for how quickly the camera reacts to large or fast-changing light levels, and how well it recovers and settles. A poor implementation can yield a complete loss of usable video for five to ten seconds while the camera tries to recover to a proper exposure level. More of the slow shutter is not always better. We have found in scenes that include motion anything slower than one-third to a quarter of a second starts to be counterproductive due to image blur. At half a second someone can actually run through the scene and not be seen.

Digital Noise Reduction (DNR): This is a feature that can help improve a low-light image by removing (or masking) some of the luminance and chroma noise present in a low-light image. It does not actually help increase low-light sensitivity, like some of the other features mentioned, but it can improve object recognition and help reduce encoder, bandwidth and storage needs. This, too, can have adverse side effects on moving objects like image smear, image softening and even image stutter. As with slow shutter, not all implementations are equal so it's recommended to compare cameras and evaluate overall benefit versus side effects. Users should also pay attention to whether DNR affects performance in bright lighting - preferably it should not.

Tinted Bubbles: This is an obvious but often overlooked item when projects are bid. Many cameras today are shipping in mini-dome configurations that include a housing and protective bubble. If low-light performance is paramount for certain locations, try to make sure tinted or smoked bubbles are not used. Depending on materials they can be 1 to 2 f-stops darker than a clear bubble. For every f-stop you lose half of your total light. This means the amount of light getting to the sensor would be reduced by 50 per cent (1 f-stop) to 75pc (2 f-stops) just because of the bubble.

High Profile H.264 compression: There are multiple profiles of H.264 and users should be wary of which profile a manufacturer uses. High profile is generally more processor intensive and is a bit harder for a manufacturer to deliver well. But the result is worth the development effort. Buyers should be aware of the ‘base profile’ H.264 implementations. Whilst the high profile is used for media such as Blu-ray, base profile is used for applications like teleconferencing. The image quality standards between those two profiles speak for themselves. For applications at the very bottom of the low-light range, look for a camera that supplies its own IR illumination.


Camera shoot-outs

We encourage everyone who is concerned about low-light performance to line up their camera options where they’re intended to be used and test them out against each other. Camera shoot-outs should be the norm for customers who care about image quality in any conditions, and this is especially true for low-light applications. At the end of the day, this is the best way to ensure that the best technology will be chosen for the application.

This article first appeared in Professional Security Magazine. Contact us if you are interested in finding out more about Professional Security Magazine or subscribing to it.