Modern surveillance cameras include ever higher levels of functionality, and the ‘headline’ features tend to be smart detection and tracking, dynamic encoding and ease of integration with third party applications. As beneficial as these functions can be, one area of performance that has been enhanced by increases in processing power – and one that is essential in many security applications – is wide dynamic range (WDR). Benchmark looks at some of the industry’s popular choices when it comes to managing hostile lighting conditions to see if they deliver balanced images.
The capture of detailed images in applications with unbalanced illumination has always been a problem for surveillance cameras. If there is a need to view an external doorway in a shop to see who is entering, or if it is necessary to cover both the internal and external views of a glazed area, without credible WDR there would have to be some compromise.
If a retailer had a camera facing a door into the premises, they would have to decide whether they wanted to see the shop area, losing details of the individual entering, or vice versa. If the camera was configured to view the shop, the bright light from outside would lead to white-out when someone entered, effectively making it very difficult to identify an individual accessing the premises.
If, on the other hand, the camera was configured to handle the bright daylight in order to identify those entering, the interior of the shop would be gloomy, noisy and often very dark.
The human eye can render images in challenging situations, but it is a marvellously complex piece of engineering. Cameras, on the other hand, struggle with any drastic illumination variance.
The issue of viewing both dark and bright areas in one scene is addressed by wide dynamic range functionality. In general, there are two approaches. The first is to capture separate images at different shutter speeds to capture dark and bright areas. These images are then combined to form one complete image.
The second approach is to sample pixels within the image, adjusting the configuration ‘on the fly’ and building a well-balanced image.
Typically, even with WDR applied, some cameras tend to favour viewing either bright areas in darker scenes or viewing darker areas in brighter scenes. Another consideration is that when deploying WDR images can display tear – a jagged artefacting – where dark and bright areas meet.
Axis: WDR Forensic Capture
Axis Communications incorporates WDR functionality on many of its cameras. Some of these have separate modes with WDR enabled which are limited to 30fps rather than 60fps which can be achieved if WDR is not deployed.
Axis offers different implementations of WDR: Forensic WDR, WDR Forensic Capture, WDR Dynamic Contrast and WDR Dynamic Capture.
Forensic WDR and WDR Forensic Capture use both dual exposures and contrast enhancement to boost detail while reducing any artefacts.
WDR Dynamic Capture is a dual exposure implementation, but can create artefacts in certain conditions, while WDR Dynamic Contrast uses only contrast enhancement which reduces the level of artefacts but has a lower dynamic range.
It is worth noting that not all WDR cameras include all variants; Forensic Capture is one of the more commonly used methods and this was used for the Benchmark test.
Configuring the WDR is very straightforward. It’s a simple case of turning it on! Obviously, there are a number of configurations for exposure which will have some bearing on the WDR performance, but these are no different to the configurations typically used when optimising any camera for best performance in a given location.
In typical applications, the WDR performance is good with clean and detailed images in both dark and light portions of the viewed scene. WDR does create certain types of artefacts, but none of these were significant and therefore did not impact on general video quality.
If you want to split hairs, there is a slight increase in noise in darker portions of the image, but you do have to be looking for it. It does not impact on the usability of the footage.
Introducing fast motion to the scene does introduce some degree of artefacting, and the severity of this is dependent upon the speed of the motion and the portion of the image it occurs in. For example, the impact was greater in brighter parts of the scene, where we saw some slight blur and ghosting on object edges. However, this was not significant enough to make the video unusable.
Very fast motion such as thrown objects introduced some outline edges as the dual frames were slower than the object. However, this is an extreme case and one that most multi-exposure WDR devices would struggle with.
All in all, WDR Forensic Capture is a good WDR implementation and in most mainstream applications it will deliver the required degree of performance without any detrimental artefacts or image degradation.
Hanwha Techwin: Wisenet X
Hanwha Techwin’s Wisenet X series of surveillance cameras make use of the latest Wisenet 5 chipset which utilises a high level of processing power to deliver advanced functionality. One of the benefits on offer from the range of mainstream cameras is WDR performance of up to 150dB (this falls to 120dB with the 5MP models).
While the majority of WDR-enabled video surveillance cameras make use of dual frame capture with varying exposures, with both images then used to create a balanced picture, the Wisenet X series uses four frames of video. Hanwha claims that the result of this is an image that has a more natural look. Another claimed benefit is that motion blur is reduced in the new WDR implementation.
The WDR can be adjusted in terms of the level (Low, Middle or High) and can be automatically disabled in low light conditions or when the device switches to night time mode (monochrome).
The Wisenet X devices also include SSDR. This can be used to increase detail in darker areas of the image if required.
The WDR set-up is found in the Camera Settings menu of the Video and Audio configurations. It is located in the Backlight section. The options for backlight management are BLC, HLC or WDR. Selecting the latter brings up the WDR-specific menus.
The WDR level can be selected from a drop-down menu. There are three options: Low, Middle or High. We didn’t see too much difference between these and all worked well, so we kept the camera set on Low.
The other options are simple on/off tick boxes for Auto (the camera detects when to use WDR), or for WDR to be disabled in low light or night mode.
WDR performance is good, and in normal conditions there’s little obvious noise or image degradation when it is being used. As the difference between darker and brighter areas increases, you do see some signs of noise, but you have to be looking for it.
Motion blur is minimal, but with fast motion it is detectable but not to a degree that impacts on the usability of the video. Very fast motion did not induce any artefacting and the image remained stable.
Darker parts of the image do drop a little bit of definition as the WDR is pushed towards its limits, but again there’s always enough visual information to ensure the video can serve its purpose.
Hikvision implements the principles of local tone mapping in its WDR. It implements global tone mapping which re-maps pixel brightness based upon the brightness and contrast of the scene. This allows image details to be enhanced whilst also retaining a more realistic look, according to the manufacturer
By exploiting the full dynamic range of the sensor, the camera remaps high bit data to 12-bit data to deliver balanced images.
The WDR is applied automatically, allowing the camera to compensate for changing scenes.
As with the Hanwha Techwin camera, the wide dynamic range configurations are found in the Backlight menu. There is a simple on/off selection for the function and a slider for WDR level (0 to 100). That’s all there is to it, so set-up is swift.
WDR performance is good once you have found the correct level for the given scene. This is achieved via a slider. If set too high the camera will over-process and the image looks a little over-brightened and artefacting is increased. Setting it too low will result in a more noisy image. For our test scene a setting of around 40 (the scale is 0 to 100) was right.
If the difference between darker and brighter areas is significant you will see a slight degree of noise, but it isn’t enough to degrade the image.
Motion blur is not an issue with typical scene movement. However, the introduction of fast motion does create some edge blur. This is visually detectable but not to a degree that impacts on the credibility of the video.
Very fast motion does deliver some artefacting, especially in the brighter areas of the scene. This is reminiscent of the artefacting shown by Axis’ WDR Forensic Capture, in that edges tend to be ‘embossed’ and there are some patchy artefacts.
Gentle tweaking of the WDR can minimise the impact of this, and a balance needs to be achieved to deliver detail and a relatively artefact-free image.
Axis Communications’ WDR Forensic Capture is simple to implement and certainly delivers the expected level of performance.
Artefacting and image degradation are minimal, as is noise. If pushed very hard, you do start to see some detrimental impact in the viewed scene. However, this occurs when the illumination is very unbalanced, and often in such cases another solution will be best practice.
As a result, it is recommended.
Hanwha Techwin’s Wisenet X cameras deliver WDR performance of up to 150dB. By making use of four frames of video instead of the more typical two, the camera does deliver a good degree of balance in harshly lit scenes.
The WDR implementation does tend to favour brighter portions of the image, but never to the extent that detail is lost in darker portions. Also, when WDR is run alongside SSDR, we didn’t see any negative impact.
As such, it is recommended.
The DS-2CD2385FWD-I from Hikvision has a good range of WDR adjustment, and we found that heading towards the lower end of the scale delivered good balanced scenes with minimal artefacts.
When high speed motion is introduced, you will see some outlining and patchy artefacting, especially in the brighter parts of the scene. However, this can be reduced by adjusting the level.
As such, it achieves recommended status.