Noise can be a significant issue with regard to video performance, impacting on quality, bitrate and storage requirements. It represents a challenge for any surveillance system operating around the clock, but reducing noise is not difficult if the right equipment is deployed. Benchmark assesses noise reduction techniques to see whether they deliver on the promise of ‘cleaner’ video footage.
Noise is a challenge for video surveillance systems, and whilst the increase in processing power for cameras and other devices means that technologies can be deployed to reduce its impact, the drive towards ever higher resolutions means that image noise is not going to go away. Noise not only affects the image quality, but it also increases required bitrate and storage capacities.
The core cause of noise is very simple. Not enough light is falling onto the image sensor. As a result, the video signal has to be amplified to increase the strength of the required image data. Of course, that also means that any data detrimental to the image caused by a lack of light will also be amplified. The result is image noise.
There are many elements that impact on the level of noise: light levels, the choice of optics, sensor size, gain levels, day/night switching points, etc.. Reducing the level of noise can be achieved, often via a combination of steps that address the above issues.
A question of compromise
There are a number of compromises often made by users of video surveillance. Some are willing to accept lower resolutions on certain cameras, because others in the system are capturing HD footage which can used for identification. Low resolution images can therefore be used to show continuity. Lower resolutions, if correctly deployed, do not affect the evidential credibility of a surveillance system.
Another compromise often accepted is reduced frame rates. Any specification of HD or 4K UHD video must be real-time to meet the standard. However, if a user is happy with 20 or even 15fps, this often will be fine (dependent upon the viewed scene) as typically any important activity will still be captured. Like lower resolutions, this might not be ideal, but if correctly deployed it would not impact on the evidential credibility of a surveillance system.
However, it is fair to say that no one has ever willingly accepted noisy video images. The push is, and has always been, to enhance video quality, not to accept poorer images. Not only can noise deteriorate image quality to a point where legal arguments can be made that evidence provided by video is not credible, but it can also impact on the overall performance of a surveillance system. It’s a compromise that end users will not accept, and neither should installers or integrators.
Whilst there are many way to approach noise reduction, including (but not limited to) the specification of cameras with larger sensors, using higher quality lenses, applying noise reduction technologies and judicious use of gain control, the two most effective measures attack the problem at its very core: insufficient light levels.
Noise is caused because of processing techniques which are used to amplify the signal, which includes video data along with any electrical noise, when not enough light has fallen onto the image sensor. If more light is added to the scene, this counters the problem. It may not always be possible to increase scene illumination, but the simplest way to reduce noise is to add more light.
Where light pollution isn’t an issue, white light can allow colour images to be retained which is a significant bonus with surveillance. Colour is a basic intuitive recognition element for humans, and so the importance of colour in a security solution can be vital.
An operator seeing a red car in a monochrome scene will see it as grey: the same colour as a blue, green, brown or grey car. This means that an important piece of information has been lost.
When used in a triggered application, white light can also be used to make any events obvious to passers-by and onlookers. It can also help with site management issues during the winter months when an application may be open for business during the hours of darkness.
If light pollution is a concern, or where a more discreet solution is preferred, infrared illumination offers a viable choice. While the industry is seeing a growing number of cameras with integral illuminators, there are many arguments in favour of standalone options.
Standalone illuminators offer more flexibility when it comes to the location of the light source. It’s not always the best option to have the light on the same axis as the camera. Also, standalone devices offer a wider choice when it comes to coverage angles and range. Integral illuminators are typically specified with restrictions to size, power and adjustability; they are also built to a cost. Dedicated illuminators can be selected for each individual camera view.
Many standalone illuminators are simple to install. Most manufacturers offer PoE versions of their illuminators, and with a number of cameras offering RJ45 power outputs, connecting an illuminator via an edge device makes financial sense.
Adding illumination is also a very cost-effective solution if you consider the price/performance ratio. As with any performance-based developments, changes at the lower end of the curve inevitably deliver a decent boost for an acceptable cost. However, as you get closer to the top limit of performance, incremental changes can carry increasingly high prices.
The other simple approach to reducing noise is to switch the day/night modes earlier. This might sound simple enough, but with some cameras it’s simply not possible.
Whilst some cameras deliver a wide range of adjustment for the day/night switching point, or allow switching to be triggered externally, many do not. During Benchmark tests we have seen cameras that offer 100 levels for day/night switching adjustment, but these equate to around 2 lux in actual use. Given the range of different requirements, even in one site, this simply isn’t enough.
Part of the problem could stem from the way that camera sensitivity is quoted. When the team at Benchmark first started testing cameras a few decades ago, low light investigations showed that most devices had a sensitivity level of around 6-8 lux. This wasn’t surprising as most manufacturers quoted such figures at the time.
As the years passed, the specified figures dropped, and today it is not unusual to see camera specifications with a sensitivity figure that is below 1 lux. Indeed, some cameras’ specifications are into three decimal places, such as 0.001 lux!
A drop in specifications from 8 lux to 0.001 lux might be viewed as a sign of great innovation and developments in low light technology. However, that’s not the case.
The main reason for these reductions is because manufacturers now measure sensitivity in different ways. Whilst there have been advances with regard to processing, they are nowhere near as dramatic as the figures indicate. For the installer and integrator, it makes it very difficult when trying to ascertain the likely low light performance in the field.
As image resolutions increase, so working in low light becomes more of a challenge. The higher number of pixels on an HD or megapixel chipset means that each image element is considerably smaller than on a standard resolution sensor. Getting the appropriate amount of light to fall onto significantly smaller pixels creates difficulties. It’s basic physics!
The Benchmark test team has often debated why some very high quality cameras are let down by having day/night switching points which are too low for many applications. The resulting noise could easily be reduced if it was possible to switch at a higher light level. One explanation might be that if a camera is advertised as having a sensitivity of 0.001 lux, it’s as good as admitting the video won’t be usable at that level if day/night switching has a range of between 10 and 2 lux.
Many cameras have a switching range of 2 or 1 lux down to 0.5 lux or below, which inevitably means that there will be noise in the images prior to night mode kicking in.
It has to remembered that noise is a result of amplification, which is necessary because not enough light is falling on the camera’s sensor to create a strong signal. It therefore stands to reason that any measure which enables a greater level of light to be sensed will help reduce noise.
It is often stated that the low light performance of higher resolution devices will not be as good as standard definition cameras. The reason has to do with pixel size and density. Image sensors require light to fall onto individual pixels in order to create the charge which sets the value for each individual picture element. If you consider a 1/3 inch SD camera chip, it contains around 400,000 pixels: the surface of the chip is divided up into 400,000 picture elements. The pixels have separate elements for red, green and blue, and the light intensity sensed is used to establish the relevant value.
An HD camera with a similar sized 1/3 inch sensor has around 2 million pixels, and a 4K UHD camera has over 8 million picture elements. Because the chip is still 1/3 inch, the pixels are so much smaller. As a result, getting the required level of light to fall on each element is a challenge.
Obviously, a camera with a larger sensor will offer enhanced performance in low light conditions, because the individual picture elements will be larger. This means they will receive a higher degree of light intensity.
Many higher resolution cameras offer a wide variety in chip sizes, with cameras using 1/2 inch or 1/1.9 inch sensors becoming more common. If these cameras are deployed with high quality lenses, the challenge of getting light to fall on the pixels is made less onerous.
The lens needs to considered in the mix as well. The lens focuses the light onto the sensor, so it stands to reason that a lower quality lens will not deliver optimum performance. Some installers and integrators baulk at the cost of higher end lenses, but the reality is that lenses with quality grinding and well designed aspherical profiles can deliver clean and focused light right up the edge of the sensor.
If you consider the cost of good quality lenses against the addition of illuminators or better quality cameras, they do offer a cost-effective boost to low light reduction.
Processing power in edge devices has increased significantly in recent times, enabling the development of enhanced features. Most credible manufacturers will include functions designed to reduce noise. This is not only because it degrades image quality, but also because of the detrimental effect it has on other performance aspects such as bandwidth needs and storage.
Predictive image compression looks for change in an image, and because noise is seen by compression engines as change within the image, it significantly increases bandwidth and storage needs. This often equates to a higher cost-of-ownership.
It is a rare camera which does not include noise reduction, and whilst manufacturers have tweaked their methods, the bottom line is that the reduction of noise is dependent upon a number of criteria. It is simple to set up a camera in a sympathetic environment to show clean images in low light. Translate that into the real-world, and problems can surface.
Noise reduction is achieved through processing of the video image, and as such the actual reduction itself can have a detrimental effect on quality. Where a camera requires other processing such as WDR, highlight compression or BLC, there will be a need for balance between all processing elements. In Benchmark camera tests, we have often seen image degradation caused by certain combinations of video processing.
Heavy-handed use of noise reduction can create smear and motion blur. Often the image will be slightly softened and then sharpened to blend out the noise; overdoing this can cause unwanted side effects.
Noise reduction techniques are improving, but for the forseeable future will serve a role in taking the edge off the impact of noise. They are certainly not a panacea for the problem!
There are two considerations with regard to noise reduction: how the bit-rate is affected, and how the image looks. As with anything, balance is key. The reduction of noise is only acceptable if the processing doesn’t introduce other degradation. Replacing noise problems with motion blur or artefacting doesn’t do anyone any favours!
Video noise certainly does impact on image quality in low light situations. Not only does it effect the level of discernable detail, which is vital if video is captured for evidential purposes, but it also is a serious negative point when it comes to user expectations.
This is especially true where a customer has invested in a system specified to deliver HD or 4K UHD video. With such systems, the expectation of quality and clarity in footage is greatly elevated. Noise is inevitably linked in the minds of many with low cost underperforming systems.
Noise does also impact on the proficiency of compression, and the degree to which this happens is quite significant and often misunderstood. Obviously, the heavier the level of compression applied, the more significant the impact will be. Whilst most applications will use constant bit-rates, thus avoiding network flooding, the degradation of the image will be detrimental to the overall performance of the entire system.
The best way to eliminate noise is to add more light! As illumination levels rise, noise falls. If the camera sensor has the right levels of light falling on its pixels, the need for amplification, and the noise it creates, is removed.
The next best option is to ensure that cameras can switch from day to night mode before noise becomes an issue. Sensitivity figures are often set for a low level video signal, and even claims of ‘usable video’ cannot be relied upon. Often, field tests are the only way to establish is a camera will deliver what is expected.
If conditions are harsh, then larger sensors and high quality optics can help.
Installers and integrators should realise that noise reduction technologies do what they state, with the emphasis on reduction. They are not noise removers! Take any ‘reduction’ too far, and other issues with image quality will result.
Noise is a bigger problem than many think, and it is also a costly issue with regard to the drain on resources. It is staggering how differently some cameras behave in adequate light and low light, and the real performance-killer is usually noise.
The best way to ensure the highest level of performance is to design systems that do not create noisy images in the first place!