Introduction to codec comparison
Codec comparisons are pretty common on the internet … Long and not very, emotional and not very, clearly skewed and not very. But most of the authors of such comparisons simply cite images from two movies side by side, as if by conspiracy to ignore the following things:
Any codec provides a different quality for frames in the same movie.
Different codecs are “sharpened” for different types of movies.
The compression quality of a particular movie can be highly dependent on the encoding parameters.
More details about each of the points:
1. Any codec offers a different quality for the frames of the same movie.
This is due to many factors. First, the codec uses a mechanism like bit rate control, which provides quality fluctuations even for good codecs. Second, the user himself chooses different bit rate strategies, and in the case of choosing CBR (or constant bit rate), the quality will be high in slow scenes and low in fast scenes. Third, the codecs have so-called. keyframes, whose quality usually changes separately and differs from the quality of the rest of the frames. Fourth, quality is affected by pre-filtering (which all modern codecs have) … Fifth … Sixth … Seventh … 🙂
This means that in any movie long enough (and the average movie is 150,000 to 200,000 frames), you can choose good and bad enough frames. Especially if one pass CBR compression was used on a fairly dynamic movie.
Those. having a long movie, codecs A and B (which are roughly equal in quality), and simple means of automating quality comparison (or a lot of religious devotion to your codec :), you can easily demonstrate with examples that A is noticeably better than B (“look at these frames!”), and that B is notably better than A (“look at these (other) frames!”). 🙂 What is being done “successfully” in large quantities. Both in press releases (a holy cause!) And on video processing sites.
The graph shows the quality in terms of PSNR (the higher the quality).
If comparing “spiky” frames, VP will be better than DivX, otherwise vice versa.
2. Different codecs are “sharpened” for different types of movies.
Many people know that DivX 4.1 LowMotion and DivX 4.1 FastMotion existed at some point. Although they played movies in the same way, their internal parameters were “fine-tuned” by one for weak movements and the second for strong ones. As a result, they harvested films of markedly different quality. Notice again: the FORMAT, in which they harvested, was one! And the compression was carried out in different ways! As a result, for the same file size, their quality is different. A similar “customization” for “your” data type is characteristic of absolutely all codecs. Someone better push high bit rates. Some are on the contrary low. Someone sharpens with a strong movement, someone underneath a weak one. And yet someone is getting better for cartoons and someone for real video. Someone under “noisy” movies, someone under “clean” movies, and so on. Having the source texts of the codec, it is possible without changing its format, that is, without changing the decompressor, sharpen it for almost any type of film. And this is used successfully in practice. This means that by having codecs A and B (of roughly the same quality), you can easily find movies where A will be noticeably better than B and where B will be noticeably better than A. At the same time, it’s funny that, formally, there may be a movie, just a DVD copy and a pirated “rag” copy taken from a theater screen can give noticeably different results in the two codecs. Those. the movie will be “one”, but the “best” (world) codecs will vary. 🙂 Having the source texts of the codec, it is possible without changing its format, ie without changing the decompressor, sharpen it for almost any type of film. And this is used successfully in practice. This means that by having codecs A and B (of roughly the same quality), you can easily find movies where A will be noticeably better than B and where B will be noticeably better than A. At the same time, it’s funny that, formally, there may be a movie, just a DVD copy and a pirated “rag” copy taken from a theater screen can give noticeably different results in the two codecs. Those. the movie will be “one”, but the “best” (world) codecs will vary. 🙂 Having the source texts of the codec, it is possible without changing its format.
