What 4K crops, muddy videos and overheating have in common
In the case of videos, manufacturers use tricks to reduce the high data volumes. The quality suffers as a result. Choosing a higher bit rate is often useless, as the problem starts when the sensor is read out.
Manufacturers have already been criticised several times because their cameras can only record 4K with a limited image section. This is of course impractical - and the "4K crop" is correspondingly unpopular. Nevertheless, cameras with 4K crop are still coming onto the market today. As no manufacturer likes to annoy their customers, the question arises as to why.
The 4K crop is part of a broader problem. Namely, that a camera has to bring the resolution of at least 20 megapixels down to a much smaller resolution. And that no method is perfect for this.
The starting point
If you record a video with a photo camera, the resolution of the sensor does not correspond to the video resolution. The sensor of the Sony Alpha 6400, for example, has 6000×4000 pixels, i.e. 24 megapixels. For videos in UHD you need 3840×2160 or in Full HD only 1920×1080 pixels. That's only around 8 and 2 megapixels respectively.
The obvious solution: the camera takes the original image and downscales the resolution - just like in Photoshop. Some cameras actually do this. This is called oversampling because the camera takes more data than it needs for the end result.
Many cameras, however, do not use oversampling at all or only partially. The problem is that the camera has to convert at least 24 images per second - in real time. Better 50 or 60 frames per second. At 60 frames, that's 3600 frames per minute that the camera has to recalculate. For a 20-minute video, this results in 72,000 images.
This requires quite intensive computing. Today's processors can do this, but they work up a sweat. They get hotter every minute. Many cameras limit video recording to protect the electronics from heat stroke.
Cameras that offer particularly high video quality should actually be cooled in the same way as PCs. However, the heat dissipation of these devices is not at its best. The Panasonic Lumix S1H is the only camera I know of that has a built-in fan. It is no coincidence that it is also an absolute top shot in the video sector.
Cropping, skipping and binning
Intensive computing processes in the camera have another disadvantage: the battery is empty within a short time. It is therefore hardly surprising that cameras use other methods to generate a resolution suitable for video in real time. These methods require less computing power, but do not achieve the same image quality.
The worst but simplest method for achieving the desired number of pixels is cropping, as mentioned above. Crop means to cut to size. With the infamous 4K crop, the image is simply limited to the required number of pixels and the border around it is cut away.
This is of course bad, because you have a different image section for photos than for videos. This makes wide-angle shots in videos impossible. These are popular in vlogging because they show a lot of the surroundings and have a great depth of field.
Line skipping is much better because it preserves the image detail. With this method, entire rows or columns are omitted from the pixel grid. This reduces the amount of data that the image processor has to process from the outset.
However, the image quality suffers because not all the image information that the sensor would provide is utilised. If all pixels are read first and a UHD image is calculated from this, the image is clearly sharper.
In pixel binning, several pixels - usually four - are combined into one. Logically, the resolution and the amount of data to be processed is then only a quarter.
Neither cropping nor line skipping nor pixel binning achieve the sharpness that results from downsampling the full resolution.
Clearly visible difference
The manufacturers do not specify which process they use for which resolution. However, if a camera delivers a crisp 4K image but Full HD is muddy, this suggests that oversampling is used for 4K and either pixel binning or line skipping is used for Full HD.
In any case, the difference in quality between 4K and Full HD is striking on our Sony Alpha 6400 editorial camera. I briefly filmed the inner courtyard of the office building in 4K and FHD, each with the highest available quality. In the image below, you can see a section with FHD on the left and a 4K section downsampled to the same size on the right.
Full HD is not always Full HD
If a camera is basically capable of downsampling the full number of pixels live to 4K, why doesn't it do the same for Full HD?
As long as 4K is limited, for example to 30 frames per second or to a certain length such as 10 or 20 minutes, other recording options are needed that do not have these limitations. Full HD serves this purpose. As a rule, Full HD can be recorded without a time limit, with frame rates of up to 120 fps. This is only possible with the "cheap" methods.
Nevertheless, I don't quite understand why there isn't a high quality option for Full HD that uses oversampling. Cameras such as the Panasonic GH5 show that it is technically possible: it calculates normal Full HD from the full number of pixels and only resorts to a less computationally intensive method at high frame rates such as 120 fps.
This means that Full HD delivers very different image quality depending on the manufacturer and camera.
A big difference in quality that is not disclosed
Specifications such as Full HD or 4K only say something about the nominal resolution, not about the actual sharpness that a camera achieves. This depends on how much information is used to calculate the video image. Oversampling utilises all the information from the sensor and is clearly the best method, but also requires a lot of computing power and can bring a camera to the boil.
I find it very irritating that there are apparently cameras that use oversampling for 4K, but a "cheaper" method for Full HD. This cannot be seen in the specifications, but it can be seen in the resulting quality. I will therefore compare the image of Full HD and 4K in future tests.
The image processors of the cameras are getting better and better. I assume that future generations will no longer rely on tricks such as line skipping or cropping, but will also be able to calculate high frame rates live from the entire sensor. <p
My interest in IT and writing landed me in tech journalism early on (2000). I want to know how we can use technology without being used. Outside of the office, I’m a keen musician who makes up for lacking talent with excessive enthusiasm.