Shulou(Shulou.com)11/24 Report--

When introducing the video shooting function, we often see terms such as point-to-point, undersampling, oversampling, pixel merging and so on. Today we are going to talk about the differences between these sampling methods.

Peer-to-peer Canon EOS R5 has about 45 million effective pixels (maximum photo size 8192 x 5464, with a scale of 3:2). When shooting DCI 8K (8192 × 4320) video, the sensor pixels are the same as the recording resolution, which is point-to-point sampling.

Similarly, 4K (3840 × 2160) point-to-point sampling requires at least 8.29 million pixels, and full HD (1920 × 1080) point-to-point sampling requires at least 2.7 million pixels. Since the vast majority of sensors are 3:2 or 4:3 instead of 16:9 or 17:9, the actual sensor pixels are a little more.

Undersampling if we turn on digital anti-shake when shooting DCI 8K video with Canon EOS R5, the edge of the picture will be cut off (the view angle is slightly narrowed). This means that the effective pixels of the sensor are reduced, and the DCI 8K resolution needs to be amplified by interpolation. The sensor effective pixel is less than the recording resolution, which is called "undersampling".

In addition to digital anti-shake, cutting, digital zoom and high-speed video shooting may also lead to undersampling. For example, in the Soni Heika camera / ZV-1 with a 1-inch stack sensor, in HFR ultra-slow motion video mode, the number of effective pixels read by the sensor decreases as the frame rate increases. When 960fps / 1000fps, the effective pixels are only 1244 × 420 or 912 × 308.

Take Canon EOS R5 as an example for oversampling. When shooting uncropped DCI 4K (4096 × 2160) video, if the 4K HQ option is turned on, the sensor still outputs DCI 8K data, which is then sent to the processor for "thumbnail" processing. The effective pixel of this kind of sensor is greater than the recording resolution, which is called "oversampling".

* in oversampling, the effective pixel of the sensor can be any multiple of the recording resolution.

The advantage of oversampling is that it can make use of all pixels, has the best detail and signal-to-noise ratio, and has less moire and pseudo-color. But "thumbnail" processing also means a larger amount of data operation, which will bring additional energy consumption and heat. And EOS R5 has a maximum of 30p when shooting 8K, so the oversampled 4K video cannot exceed 30p.

Skip another way for high-pixel sensors to record lower-pixel video is "skip mining", that is, skip some pixels. Taking EOS R5 as an example, we can only read odd lines on the sensor to get 8192 × 2160, and then get DCI 4K video from horizontal thumbnails.

Because some lines are skipped, it takes less time to complete a read. It can not only achieve a higher frame rate (50p / 60p), but also achieve a lower shutter shutter (less jelly effect). However, because the actual use of fewer pixels, so the signal-to-noise ratio and resolution will be significantly reduced.

* both jump mining and cutting give up some pixels. The difference is that jump mining uses a larger sensor area, which usually does not affect the shooting angle and visualization.

Pixel merging in addition to oversampling and jump mining, some new sensors also support "pixel merging", that is, the information of adjacent pixels is merged and then output to the processor. This not only reduces the pressure on the processor, improves the reading speed, but also uses all the pixels to achieve a higher signal-to-noise ratio (the effective area of the sensor remains the same, but the resolution is not as good as oversampling), which is a better working mode than skip mining.

The four Bayer and nine Bayer sensors on the mobile phone perform a "pixel merging" operation during daily shooting, and the effective pixel of the sensor is an integral multiple of the recording resolution.

Oversampling and full sampling as mentioned earlier, the DCI 4K video obtained by 8.2k oversampling uses all the pixels in the sensor at 17:9. This extreme oversampling can also be called "full sampling". Oversampling is not the only case of "full sampling". In many cases, it also occurs at the same time as interlacing and pixel merging. For example:

When shooting uncropped 4K video, Sony α 1 first merged from 8.6K pixels to 4.3K, and then oversampled to 4K.

Canon EOS R5 first jumps to 4K and then overtakes full HD video when shooting uncut full HD video.

This article comes from the official account of Wechat: camera Notes (ID:xjbiji), by Teng Fei et

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.