
How does bitrate affect video quality?

To understand how bitrate affects video quality, we first need to understand how video compression works.

When you’re compressing video, the compression algorithm does two things:
Convert video from “pixel domain” to “frequency domain” by DCT (Discrete Cosine Transform). If you are not familiar with DCT, you can read our previous article: How to explain DCT to a 5-year-old?
Some of the information in the frequency domain is discarded through a technique called quantization, with the expectation that the human eye will not perceive this loss of information.
Essentially, when you compress a video, you’re throwing away some information while hoping there’s no loss in video quality. When you overcompress the video, you throw away a lot of information, and the human eye perceives the encoding loss. If you don’t compress the video too much, the file size will be huge, while the video quality will be excellent.
This is known as the “rate-distortion exchange (RDT)” in video compression, and the lower the bitrate, the worse the video quality (assuming the resolution is fixed).
What does RTD mean?
Let’s look at the example below. We use FFmpeg to compress the example video (CrowdRun) to 1.5 mbps and 5 mbps, respectively. With the video screenshots below, can you tell me which one is from the higher bitrate video and which one is from the lower bitrate video?
Obviously the bitrate in the image above is lower and the bitrate in the image below is higher. When using a bitrate of 5 mbps @ 1080p, the video quality is excellent. For the same video, when we use a bitrate of 1.5 mbps, the video quality of CrowdRun becomes very poor.
But is it always like this? Does low bitrate always produce poor quality video? We answer this question with a simple experiment.
Let’s test the “low bitrate = low quality” hypothesis using animated videos. For the following video, we used the popular Simpsons trailer and compressed the video to 2.5mbps and 1mbps using H.264/AVC (make sure all encoder settings are consistent).



