CCTV Test: The Truth about Sensitivity
Sensitivity is a specification that should tell installers and integrators a lot about a camera’s performance. Indeed, it was once the case that those seeking cameras would look first at the resolution and sensitivity specifications, and often those two alone would significantly affect the purchasing decision. Benchmark considers whether the camera sensitivity specification is still credible in today’s market.
Axis Communications – Q1615
The AXIS Q1615 is an HD1080p camera which delivers streams at up to 50fps. The camera uses a 1/2.8 inch CMOS sensor, and video is compressed using H.264 (baseline, main or high profile) or Motion-JPEG. Multiple streaming is supported. Sensitivity is claimed to be 0.36 lux (F1.3) at 50fps, or 0.18 lux (F1.3) at 25fps. The camera includes the manufacturer’s Lightfinder technology.
The camera is supplied with a 2.8-8mm varifocal IR corrected P-Iris megapixel lens from Fujinon, and has lens-specific functionality.
Bosch – Dinion IP Starlight 8000
The Dinion IP Starlight 8000 is a 5 megapixel camera, and makes use of a 1/1.8 inch CMOS sensor. It delivers real-time video footage at 5MP in 16:9 (2992 x 1680) and 4:3 (2704 X 2032) aspect ratios, along with HD1080p and HD720p streams.
The ‘Starlight’ designation indicates low light technology, and the quoted sensitivity figures are 0.0121 lux in 5MP mode, or 0.00825 lux in HD1080p mode. Both figures are quoted for a 30IRE image.
The DH-IPC-HFW8301EP is a networked day/night camera which delivers 3 megapixel, HD1080P, HD720p and standard definition streams. It utilises H.264 and M-JPEG compression, and triple streaming is supported. Sensitivity is claimed to be 0.1 lux. WDR has a 120dB range.
The camera features integral infrared LEDs for illumination; these have a smart function which adjusts power to eliminate potential white-outs.
Pelco – IME219-1ES
The IME219-1ES is a networked day/night camera which delivers HD1080p streams. It utilises H.264 and M-JPEG compression, and multiple streaming is supported. The unit includes integral IR leds for night vision. Sensitivity is quoted as 0.2 lux.
The lens is a 3-9mm varifocal unit. Features include adaptive motion detection with sabotage protection, SureVision WDR, alarm I/O, two-way audio and edge recording via a MicroSD card. Power is PoE.
Samsung Techwin – SNB-6004P
The SNB-6004P is an HD1080p box-type camera which features the WiseNet III processing engine. The camera streams at rates of up to 50ips, using H.264 and Motion-JPEG processing. The camera uses a 1/2.8 inch CMOS sensor, and has a claimed Sensitivity of 0.1 lux for a 50IRE image.
Features include support for on-board apps, SSNR, SSDR, auto-focus, edge recording, intelligent video analysis and a host of video processing features such as defog and facial detection.
[dropcap]I[/dropcap]n the past 20 years, the video surveillance sector has seen a lot of technological advances. Manufacturers have R&D teams working around the clock to deliver an ever higher level of performance to those seeking to deploy video systems. Resolutions have been raised from 330 TV lines to today’s HD streams and multi-megapixel devices, with moves towards 4K across the mainstream market and higher 7K streams being announced at the time of writing.
Resolution hasn’t been the only advance. Functionality has also increased. Previously CCTV cameras offered a few basic image manipulation functions. Additional features such as WDR, privacy masking, multi-streaming, VMD and IVA all started out as select options on top-of-the-range devices, before becoming standard features.
It is fair to say that over the years video surveillance has developed significantly in terms of performance. If you look at the specifications for products over this period of change, you’ll see a significant shift in figures quoted for Sensitivity. This specification should give an indication of a camera’s ability to deliver a video signal in low light, and as such is – or at least should be – a very important piece of information.
Unfortunately, there is no defined or standardised method for measuring Sensitivity. In the past there was an industry ‘understanding’ that Sensitivity should identify the lowest light level, measured at the viewed scene, which allowed the camera to deliver a 1 volt peak-to-peak signal with all processing switched off. This measurement was taken using an F1.2 lens.
Using this method of measurement, mainstream colour cameras typically had a Sensitivity of 8-10 lux. Today, it is not uncommon to see quoted figures of 0.001 lux.
You would be forgiven for thinking that a drop in specifications from 10 lux to 0.001 lux is a sign of great innovation, of rapid developments in technology, and therefore a cause for celebration in the surveillance sector. However, it’s not the full story.
Whilst some advances have been made with regard to low light performance, the main reason for these dramatic reductions is because manufacturers measure Sensitivity in different ways, and this can significantly impact on the final figures.
Sensitivity can be affected by a wide range of different variables: whether measurements are taken at the viewed scene or the faceplate, scene reflectance, shutter speed and frame integration, gain levels, lens aperture, video signal quality, etc..
It is not unusual to see Sensitivity figures which, when analysed, will very obviously give a final image that simply will not be suitable for surveillance. It must be remembered that Sensitivity figures do not in any way imply that the video captured at the stated levels will be fit for purpose.
As such, the specification is of little or no use. The fact that often the full variables of testing are not disclosed means that the figures cannot be used to carry out a theoretical comparison.
Where advances in Sensitivity have occurred, these have often been countered by other challenges associated with higher resolution chipsets. The majority of HD or megapixel chipsets are a similar size to standard definition image sensors. The increased number of pixels means that each image element is considerably smaller than on the standard resolution sensor. Getting an appropriate amount of light to fall onto significantly smaller pixels creates difficulties. It’s physics, and all the marketing spiel in the world can’t change that!
So, what’s the reality when it comes to Sensitivity? Whilst we have to accept that quoted figures are not indicative of the performance we will receive (for reasons explained in the panel entitled ‘The Specification Game’ below), it helps to understand what you can expect in low light applications.
It must be said that the ‘old fashioned’ way of measuring video no longer can be applied to most image capture devices. Digital data streams have replaced electronic signals, and a straw poll in the Benchmark office indicated that it has been a good few years since we’ve seen a camera which allows all processing to be disabled! Indeed, today many cameras rely on processing for a wide range of tasks. This is no bad thing when it’s applied well!
To assess the performance all cameras had day/night switching disabled to force a colour image. We sought the views of several end users and identified their requirements. Whilst they didn’t seem too worried about a slight fall in frame rate, they didn’t want jerky motion or image blur. A good degree of colour fidelity was also wanted, as was a low level of noise.
Where cameras included low light processing to boost performance, this was utilised, but only to a degree which ensured performance did not fall below the level of the end users’ expectations. All light measurements were taken at the viewed scene.
The cameras used for this analysis of Sensitivity have all been tested by Benchmark in the past, and all have a role to play in video surveillance. Whilst aimed at slightly different market sectors, all are credible video capture tools. The aim was not to consider the real-world performance against the specified figures, but to highlight what installers and integrators can reasonably expect to achieve.
The AXIS Q1615 is one of the manufacturer’s newer models and features Lightfinder technology, which is an automatic function. It is a processing function to enhance low light performance.
The Sensitivity specification for the Q1615 is quoted as 0.18 lux for a 25fps stream. With regard to configuration, there is an option to optimise the image for less noise or less motion blur.
The Lightfinder technology does enable colour images to be retained at lower light levels than with standard image processing, but there is a trade-off. For good fidelity colour information, a fair degree of processing noise is introduced. Reduction of noise inevitably generates an increasing level of motion blur.
The image is actually good down to around 4 lux, at which point you have to make some decisions. Whilst noise will increase, it is preferable to motion blur, and at around 3 lux you really are on the edge of decent quality footage. Despite this, a usable image (and we mean usable in its loosest sense) can be achieved at around 2 lux. It is noisy, and there are very slight signs of motion blur.
The Dinion IP Starlight 8000 is, like the Q1615, a camera which carries a premium price tag. It does deliver a lot of functionality, and uses a 1/1.8 inch sensor, which helps with low light performance.
A new offering from Bosch, the camera performs very well, and is arguably the first mainstream networked camera to truly eclipse the low light performance associated with top-of-the-range analogue models.
Its Sensitivity specification is 0.00825 lux. This figure is quoted for a 30IRE image, which no-one will be happy with having invested in a high end camera!
Even with a bit of rounding up, if you want usable colour, a relatively clean image and blur-free motion, 0.01 lux isn’t a realistic target. However, 2 lux will deliver all of the above, and with the right set-up the image won’t appear to be heavily processed. If you’ll accept some degradation, a usable image can be achieved in conditions of 1 lux; that’s our idea of usable!
The difference between 0.00825 lux and 2 lux might seem immense, but 2 lux for an HD stream is very good indeed in the real world.
Whilst the Axis and Bosch cameras are unashamedly high end devices where you very much get what you pay for, the DH-IPC-HFW8301EP from Dahua is a more humble option, aimed at the budget end of the market. Despite this, it weighs in with a quoted Sensitivity of 0.1 lux.
Colour fidelity suffers and visible fading becomes obvious at around 6 lux, as do other signs of image degradation. Noise is fairly prominent, although motion blur only becomes an issue in motion-heavy scenes. This isn’t aided as the higher degree of noise obviously impacts on the efficiency of the compression, and with a constant bit-rate overall performance does tend to become more sluggish.
There is something of a struggle between noise and motion blur as light levels drop further, but as there’s no effective way of managing the balancing act conditions below 5 lux do affect image quality.
Designed for ease of installation, the camera offers limited configuration options to installers and integrators. As a result, between 4–5 lux, the video footage becomes unusable for credible surveillance in colour mode.
The camera does features integral infrared LEDs. It is not uncommon for cameras equipped with additional integral lighting to struggle in lower light conditions, as the product specification is intended to appeal to those seeking IR operation.
The same could be said of the Pelco IME219-1ES, which also includes integral infrared illumination. The dome camera delivers HD1080p streams, and has a quoted Sensitivity specification of 0.2 lux. As light levels fall, colour rendition remains relatively faithful, and for those with a requirement for colour information this is a positive.
Motion remained smooth as light levels fell, and the IME219-1ES showed that it could deliver high frame rates with minimal motion blur. Frame rates remained above 20fps, and while very fast motion did have a slightly ‘processed’ look to it, the degree of blur was minimal, thus ensuring that detail was preserved.
Whilst motion and colour rendition were good, noise did become an issue, and at 6 lux it became obvious. Whilst some noise can effectively be reduced via various configurations, the trade-off is that the processing will create slight blur and shadowing in the image.
Again, it must be stressed that 5 lux is a typical point for low light issues in mainstream cameras, despite being some way off the quoted figure of 0.2 lux. Again a case could be made that if an IR-equipped device is selected, the user will probably be happy with a cleaner, albeit monochrome, image.
The SNB-6004P from Samsung Techwin uses the WiseNet III processing engine, and has a claimed Sensitivity of 0.1 lux for a 50IRE image. The WiseNet III platform has been around for a while now, but the manufacturer has continued to develop it. Recently the concept of open platform Apps was added, and as this is still at an early stage indications are that the III incarnation will stick around for a while.
Whilst Apps might be a headline feature, many of the additional functionality addresses performance in both low and poor light. The definition of the latter is uneven, unbalanced and harsh lighting. The camera includes SSLE – Samsung Super Light Enhancer – and this is similar to many of the proprietary features designed to boost low lighting.
As with the units from Axis and Bosch, the high degree of flexibility does offer the ability to balance noise and motion blur. This allows colour information and image detail to be retained until around 3 lux. It might not be a perfect image, but it’s certainly acceptable to most end users.
If you want to drop to a lower light level, you’ll have to accept some degree of obvious degradation. SSLE will boost the image, but expect fast motion to show elements of blur, and the general image has a slightly fluid look to it.
Footage remains usable to 2 lux, and some will argue 1.5 lux, but the latter split opinion in the test team!
Sensitivity figures have changed over the years, and whilst some of the quoted numbers are an obvious nonsense, don’t be surprised to see them pushed even further. Despite this, the right level of light still needs to fall onto the pixels to create a usable image, and no number-crunching will change that fact.
We would not expect to see low light performance change dramatically in the near future, and in reality the performance of HD and megapixel cameras has never been better, despite what the numbers might say!
Expect 4-6 lux from good quality mainstream cameras and 2-3 lux from premium models, and you shouldn’t be too disappointed in the real-world performance you achieve!
The Specification Game
Some might wonder why manufacturers continue to use Sensitivity specifications that do not give a clear indication of how well a camera will perform in low light. The answer lies in how specifications are written, and highlights an approach that cannot be described as best practice.
Often, specifications will be generated using a cut-and-paste approach. There are two ways this can happen. The specifier might identify a camera which he believes will fill the brief. In the interests of impartiality, he will not identify the device, but will instead cut and paste the specification for it into the tender document. Sometimes, if they can’t be bothered to do that, they might just lift the specs from a previous document with similar needs.
If the selected device specification comes from a manufacturer that quotes Sensitivity for a 30IRE image, then a similar (and maybe better performing) camera which quotes for a full video signal or even 50IRE will not seemingly meet the specification.
Because there is no defined method of measuring Sensitivity, some manufacturers will always find a way to weight the figures in their favour. If specifiers of the cut-and-paste persuasion then use those figures, it forces all manufacturers to adopt the method of measurement used by those who initially manipulated the figures. It becomes a race driven by the lowest common denominator.
The loser in all of this is the installer and integrator (and specifiers with a working knowledge of video) because what was once a very useful specification has become something of a nonsense!