Pages

Monday, November 24, 2014

To 4K or not 4K video or Ultra HD

To 4K or not to 4K video or Ultra HD
Our industry’s seemingly insatiable appetite for more and more resolution has now produced a wave of interest in 4K cameras that promise exceptional clarity and sharpness, akin to the big screen, Ultra HD television sets found in consumer electronics stores and an increasing number of North American homes.
The jury is still out on whether there is an immediate need for the resolution that can overcome the downsides of increased storage and bandwidth required for running 4K cameras in a surveillance operation. Like so many things, if the cost of the camera, cost of the supporting system infrastructure and components were of no concern, this new format would likely be a more viable and attractive option for many security applications.

Here are four things to consider before making the leap to investing in and deploying 4K video:

1. What will I get with 4K that is not possible at lower resolution?
There’s no doubt that 4K technology is light years ahead of analog quality, but the reality is that the increased clarity and sharpness provided by that level of resolution is often over and beyond what is required and able to be managed by a typical security operation. For many reasons, full HD/1080P is the most commonly used resolution for new systems. The majority of security systems in use for live monitoring situation do not really benefit from such a resolution, as the human eye is well served with the details of a 1080P picture. Higher resolutions pay out when more details are required in forensic investigations.

2. Double the resolution, double the processing requirements
Users typically want to see more than one camera on one monitor, and only  occasionally switch to full screen modes. With 4K, the clarity of that multi camera view would be no clearer than what would be viewed from a lower resolution camera. In addition, delivering streams from multiple 4K cameras presents some technical challenges. The client PC and graphics card must handle a significant flow of data. The best approach is to have the live view limited to only enough resolution for the video size and screen resolution of the display.
Today a typical approach to balance PC power requirements and quality uses lower resolution streams for live view, while recording in the highest resolutions. 4K resolution taxes the workload on the network because recording the highest resolution means the full stream content moves from the camera to the NVR.

3. Limitations on form factors, lenses
The availability of affordable high resolution optics is just not there yet, and a dome style camera with a typical curved dome bubble cannot transmit the 4K resolution. In addition, a true 8MP resolution lens with appropriate coverage for the 4K sensor is quite large, which would render a 4K version of the compact dome camera (the market’s favorite form factor) essentially not possible.  The dome camera would get physically bigger which, for many customers, is a negative.

4.  Bandwidth and storage requirements
From a cost perspective, quadrupling the resolution from full HD to 4K won’t quite double the camera price. However, on the recording side it will most definitely demand more than double the storage requirements when operating under the same conditions.

Bandwidth consumption is related to processor power available on the camera. For example, the average full HD cameras deliver about 6Mbps at 30 ips. On the bright side, some manufacturers are offering full HD models with advanced compression capabilities that can reduce bandwidth consumption to about 3Mbps, with the next iteration to handle 4K video at full HD bandwidth consumption levels. Additionally new compression standards such as H.265 HVEC (High Efficiency Video Encoding) will make higher resolution bandwidth more practical for surveillance.

So where does this leave you, 4K today or not just yet? For some customers a bigger number is frequently perceived as a better solution but surveillance installations should focus on the reason the system investment is being made in the first place; protection of personnel and protection of assets. It is far from a one size fits all decision and resolution is an important tool in the system solution.

Next benchmark for video surveillance cameras is going to be the Ultra HD standard, with a resolution of 3840 x 2160 – around 8MP. Given the challenges networks may face carrying Ultra HD video streams it’s hard to say just when we will see the technology reach a tipping point.

THERE are a couple of signs worth paying attention to with Ultra HD (commonly called 4K in consumer and CCTV industries). The first is that UHD consumer monitors, which are now dropping in price at a time many homeowners’ first 1080p HD monitors may be starting to look a little tired. Something else to bear in mind is the consensus forming in digital photography that 8MP is the sweet spot that allows the best balance of low light performance and high resolution.

As most readers know, the more pixels you cram onto an imager, the smaller those pixels must be. And the smaller the pixels, the less light they can absorb. Double the number of pixels on a 1/3-inch HD sensor and you halve the light reaching the sensor. The result is that more pixels does not a perfect camera make – not unless sensor sizes increase. If it’s all about display images today, then 3MP cameras with a 1080p resolution are ideal.

But if you need digital zoom or you use a UHD monitor, then Ultra HD cameras should be a consideration. If you zoom in 2x digital with an Ultra HD image then you are viewing at 2MP, which is pretty good considering how quickly it takes an HD camera to burrow down under 4CIF when digital zoom is applied. 

Something else to bear in mind when considering digital image quality is that pixels on a digital camera’s sensor capture light in red, blue or green – not all colours at once. A layout will be a pair of green, a red and a blue in a grid pattern and onboard software then nuts out the colour value for pixels. This means there’s signal attenuating averaging going on in the background - one colour per 4 pixels.

A camera’s digital engine is also working hard to stave off false colours and moire – spacial aliasing that causes false patterns in a scene. Camera engines will blur an image slightly then sharpen it in order to lose such artifacts. Clearly, the more pixels, the more work the camera processor has to get through and this can be noticeable as latency or blurring if there’s sudden movement – like cars moving at right angles across a scene.

Lenses are another issue. An Ultra HD camera is going to need a quality lens and there’s no doubt that plenty of 1080p cameras are being sold with lenses that are not perfect. Sure, things look good in the centre of the image but out towards the edges details get muddy, especially on the sorts of deep zooms that might motivate a buyer to choose Ultra HD in the first place.

Sensor noise is also something that has to be considered at multiple levels. In low light, cameras increase exposure, elevating noise levels. Furthermore, pixel measurements are never perfect and the flaws in these signals show up on a monitor as noise. It’s unhelpful during the day and blinding at night. Digital noise reduction is the answer but DNR processing doesn’t just lose noise, it sloughs fine detail off a scene. When you look at a camera being tested in low light you can often see the areas where DNR has scrubbed and smudged a scene free of detail. It’s not a good look.

Engineers can build high resolution imagers with fewer noise problems but they need to be big – 1-inch or 1/1.5 inch sensors are ideal. A 1/1.5-inch sensor has 4x the area of the 1/3-inch sensors that typically run inside 1080p CCTV cameras. No wonder GBO’s S1080 camera (BGWT sells them in Australia) with its monster 1-inch sensor has such a great image in low light.

If the sensors are large enough, the lenses are good enough, the network is capable enough and the storage sufficient, then Ultra HD cameras will give end users a lot more detail than 3MP 1080p cameras can. But this capability has to be balanced against many things. Image quality is about more than megapixel count. 

“If a sensor is of the similar size as the equivalent HD sensor and it has 4x the pixels - low light performance will be 4x lower. And streaming bandwidth will be close to 4x larger unless better compression is used”

Keep in mind,
4K =     8.3 megapixels, aspect ratio 16:9 Horizontal resolution
1080p= 2.1 megapixel,  aspect ratio 16:9 Vertical resolution
720p=   1.3 megapixel,  aspect ratio 16:9 Vertical resolution

D1=      0.4 megapixel,  aspect ratio   4:3 Vertical resolution

3 comments: