Camera Basics #24 - What are video codecs?

The compression of video files

Now that we have clarified what a codec does and which ones there are, let's look at the level of compression within the codec. In the following, we will call the individual images of a video frames, because a video has 24fps, i.e. 24 individual images per second, but since the term comes from the English language, we will now explain it correctly for the sake of form and according to the industry standard.

GOP (Group of Pictures) / Long GOP (Long Group of Pictures)

In GOP compression, several frames are taken together as a group. There are different types of frames within a GOP: I-Frames (Intra-Frames), P-Frames (Predictive Frames) and B-Frames (Bi-directional Predictive Frames).

Features

• Efficient compression with lower data rate and file size compared to All-I, as redundancies between the frames are utilized.
• More complex decompression, as several other frames have to be decoded to play back a B or P frame.
• Can lead to loss of picture quality if the GOP is too large or too many B and P frames are used.

Use cases

• Streaming and broadcasting where bandwidth is limited.
• Scenarios where file size and transmission efficiency are more important than ease of editing.

All-I (Intra-Frame) / All-Intra

With All-I compression, each frame of a video is compressed and saved independently of the other frames. This means that each frame is treated as a complete image.

Features

• High image quality, as each frame is available in full quality.
• Higher data rate and larger file size compared to GOP, as more data has to be stored.
• Ideal for video editing and post-production because each individual frame can be accessed directly.

Use cases

• Video editing and post-production.
• Scenarios where high image quality and editability are more important than file size.

HLG (Hybrid Log Gamma)

Hybrid Log Gamma is a type of HDR (High Dynamic Range) developed by the BBC and NHK.

Features

• Enables the recording of HDR content that looks good on both SDR (Standard Dynamic Range) displays and HDR displays.
• Often used in combination with other compression strategies.

Use cases

Particularly useful for broadcast and live transmissions where seamless compatibility with different display types is required.

RAW

RAW formats store the image data directly from the camera sensor in unprocessed form. You may be familiar with this from photography.

Features

• Offers maximum flexibility in post-processing.
• Very large file sizes, as only minimal compression is applied.

Use cases

Professional film and photo productions where maximum quality and flexibility in post-production are required.

ProRes

ProRes is a codec developed by Apple. It offers high quality and efficiency in the editing and playback of video content and is often used in professional film and TV productions. It is important to know that Apple owns a patent on the hardware-side decoding, which is why it works much more efficiently on IOS systems than on a Windows system.

Features

• ProRes offers excellent image quality at relatively moderate data rates, which makes it ideal for post-production.
• Different variants offer different degrees of compression and quality levels that are adapted to different production requirements.
• Sub-compression types: 4:2:2, 4:4:4:4 (4:4:4), LT, HQ, Proxy and RAW

Use cases

• Due to its high quality and ease of editing, ProRes is often used in professional post-production workflows.
• ProRes is widely used in film and TV production because of its efficiency and quality.
• ProRes is often used in broadcast production because it offers high image quality at moderate file sizes.
• ProRes 4:4:4:4 and ProRes RAW are particularly useful for visual effects and color grading as they offer high color depth and flexibility.

Color sampling / chroma subsampling

The human eye reacts better or more sensitively to differences in brightness (luminance) than to colors (chrominance). This is also the reason why many black and white images appear sharper to the human eye than a color image. We also go into this in our camera sensors article.

So what is behind the three numbers separated by a colon? You are probably already familiar with this information: 4:2:0 or 4:2:2, they are quite common in the various camera menus. They indicate how brightness and color information is weighted over two rows and often four columns. To explain the whole thing better, we will now transfer the information to A:B:C and try to make it easier to understand.

Important for understanding: Chroma subsampling always combines 2 rows of pixels that lie below each other

Chroma series

Our A, which is usually marked with a 4. Describes 4 pixels in the horizontal, which form a row, of which 2 fields each have the same value.

Number of color information

B describes how much color information the horizontal chroma row is composed of, so here a 4:4:X means that each pixel is assigned a color value, so if it has a 4:2:X, then only half of the color information per pixel will be stored.

Where our C describes the second row and how the colors or rather the chrominance behaves, this is 0, it copies the value of the row above.

Quick explanation

• 4:4:4 = 4 pixels of brightness information, 4 pixels with their own color information (1 row) and 4 pixels with their own color information (2 rows); a lot of computing power required

• 4:2:2 = 4 pixels of brightness information, 2 pixels with their own color information (1 row) and 2 pixels with their own color information (2 rows); a lot of computing power required

• 4:2:0 = 4 pixels brightness information, 2 pixels with own color information (1 row), 0 pixels own information; here the information of the row above is taken over; less computing power required

The whole thing can be supported by GPU acceleration and displayed better.

Which one do I use for what?

As a rule, you can hardly see the differences, but if you are working with a chromakey, i.e. the typical green or blue backgrounds, then it depends on every color information, otherwise the mask could become unclean, here values above 4:2:2 are particularly suitable.

Conclusion

It is very useful to familiarize yourself with the various video codecs and how they work, as this can help you to assess what you need for your next shoot. Or whether your own PC is capable of processing the data.