What is per title encoding?

▲ Scan the QR code in the picture to learn more about the audio and video technology conference ▲

---

translate 、 edit ：Alex

Technical review ： Zhao Jun

This article is from OTTVerse, The author is Krishna Rao Vijayanagar.

Easy-Tech #036#

Per-Title（ By theme ） code To save bit rate 、 Storage space and ABR The transmission bandwidth is for each movie （ Based on its unique spatial and temporal attributes and complexity ） adjustment ABR Rate ladder （bitrate ladder）. In other words ,Per-Title The purpose of coding is based on the characteristics of the film （ Slow motion 、 sports 、 Animation 、 Cartoon content, etc ） Generate a different set of encoding or compression parameters for each movie .

Next , We will understand Per-Title The coding process involved in coding and the advantages it brings to streaming media providers .

What is? Per-Title code ？ Where to start ？

First mentioned Per-Title The coding place is Netflix The blog of , And then IEEE Published a book called “Complexity-based consistent-quality encoding in the cloud（《 Stable quality coding based on complexity on cloud 》” The paper of . There is an interesting sentence in the abstract of this paper ：

In order to produce the best quality video stream , The system needs to adapt the coding to each content （ In an automatic 、 Scalable way ）. In this paper , We describe two Algorithm Optimizations , For cloud based distributed coding pipelines , They are ：（1） For bit rate - Resolution selection Per-Title Complexity analysis ;（2） For stable quality coding Per-chunk rate control . Compared to the simple “ One size fits all ” Coding system , These improvements have brought many advantages , Including more efficient bandwidth use and more stable video quality .

The above sentence “ In order to produce the best quality video stream , The system needs to adapt the coding to each content ” A good summary of Per-Title code .

Encoder needs “ understand ” Every video content , And adjust the compression settings and parameters to suit it , In this way, it is possible to achieve the best video quality .

| Conventional ABR And what happened to the compression process ？

In the use of ABR In the traditional video transmission method of Technology , It is generally to create a code rate ladder （ Or a group profile）, And apply it to all movies in the content library . About ABR More introduction to technology , Please read this article ： understand ABR And how it works .

such as , There is a bit rate ladder 6mbps 1080p Of profile, And applied to all classifications —— Whether it's animation 、 Sports or talk show .

However , There is a problem with this method ： The characteristics and complexity of each film are different .

All movies look different ：

Some movies have fast action scenes （ Sports competition 、 Action movies ）, Some are slower in action （《 Shawshank redemption 》）. Some cartoons are relatively simple （《 The Simpsons 》）, Some have a high degree of detail （《 Toy Story Mania 》）. All movies have their own “ gene ” And features , So every movie produced is different .

that , Why should we compress movies in the same way , Use the same encoder settings and use the same code rate ladder ABR Video transmission ？

Let's take a look at the following from 《 The Simpsons 》、 Football matches and Park Joy Three screenshots of the test sequence , They all look different , Is that so? ？

Easy to compress ！

It's really hard to compress ！

Because there is water in the video 、 Grass and leaves , It's also hard to compress ！

Now? , These examples depend on your subjective judgment about the quality of video . Let's see Netflix Digital experiments on the technology blog . At the bottom of the RD The figure describes the bit rate and video quality of different sequences at different target bit rates （PSNR）.

Look at the changes in the figure ！ stay 5000 kbps, Some sequences have up to 45 dB Even higher PSNR The score is , Other sequences only 36 dB. It makes it clear that ： No two videos are the same , They should be handled according to their respective characteristics .

In more technical terms, it is , There are differences between the spatio-temporal complexity and characteristics of these videos , So it would be a good idea to use this to compress video effectively .

source ：Netflix Blog [1]

therefore ,Per-Title Coding is changing from a video to （ Or adapt to ） Another video coding .

Use Per-Title code , Which variables can be changed ？

Use Per-Title When coding , Many coding and transmission parameters will change , such as ：

Resolution selection in the code rate ladder ： some title It might generate 720p Quality content （ It also looks great ）, For this kind of video , You may not have to switch to higher quality content 1080p.

The bit rate selected for each resolution ： This is a Per-Title The most important part of coding . If you have to generate a set of video resolutions （1080p、720p etc. ）, Then you can change the bit rate for each resolution . in other words , You will find that you may not be 6mbps Time generation 1080p In the video , But in 3mbps Generate 1080p, And achieve the same video quality ！

In the code rate ladder profile Number ： This is a Per-Title Another advantage of coding . By changing the code rate - Resolution combination , Maybe it can reduce the number of bits you need to generate in the rate ladder profile Number .

In the use of Per-Title When coding , The range of its parameters is larger . At a more subtle level , You can study the encoder settings and adjust them ：

• Strength of filter

• GOP length

• Enable and disable half pixel or quarter pixel motion estimation

• Search range of motion estimation

• GOP structure （P Frame B Frame ratio ）

And more depends on how to set up the video codec . The first priority here should be to understand your video complexity , Video codec capabilities , And how to combine all your data and video intelligent analysis to effectively compress video .

How to achieve Per-Title code ？

Per-Title The most important feature of coding is that it can “ understand ” The complexity of a movie 、 The scenes and changes . The way is ： By collecting movie information and statistical data , And use these data for compression .

This makes us need to know about multi pass coding （multi-pass encoding） The concept of , The first time （ Or N All over ） It is used to collect information about movies . On the last day M All over , Use this information to encode video .

What information is helpful to understand the complexity of the movie ？ Let's see ：

Global velocity or motion vector ： It will tell us how fast the scene moves , Can be used to distinguish talk shows （ None of them moved ） And American professional football （ Full of fast camera movement ）.

Spatial complexity ： Most of the pictures in the movie are like 《 The Simpsons 》 Solid color block in ？ It is still full of the complex patterns in the movies of the s ？

Time complexity ： Want to understand how quickly movie content switches from one frame to another , This is also related to the global motion vector and velocity in the above .

These are very important video features , They determine how to compress video effectively with a certain bit budget . In a nutshell , If you know your video properties , You can adjust the encoder settings to achieve the best video quality （ For example, it is required to compress the video to x mbps）

therefore , After you collect this information , You can execute another one on the video codec pass To compress the video to the correct bit rate （ It's up to your convex hull algorithm ）.

Per-Title The advantages of coding

perform Per-Title Coding has many advantages , such as ：

Save storage space ： By using Per-Title Coding changes the bit rate and resolution , You can compress video efficiently , And save a lot of storage space .

Saving transmission cost ： Because each is encoded title All use a rate ladder that is most suitable for it , So you will soon see CDN Transmission cost savings . besides , End users will also download smaller files , So as to reduce the occurrence of buffer and the first screen delay .

Save coding time ： also , Because the coding ladder will be adjusted separately for each film , You can easily see the savings in coding time . such as , If we don't use 1080p code 《 The Simpsons 》 Fragments of , But use 720p And get the same visual quality , Then the decrease of resolution will increase the speed of encoder . This is mainly due to the reduction of resolution, which reduces the work of motion estimation and compensation algorithms .

Improve quality ： By adjusting the encoder of each movie or theme 、 The resolution of the 、 Bit rate 、 Frame rate and other settings , You can make full use of the encoder , And get the best video quality . This will bring a great user experience ！

therefore , By switching to Per-Title coding scheme , You can save a lot of storage 、 The cost of transmission and coding time .

That's it today , See you next time , Take care ！Happy streaming!

notes ：

[1] https://netflixtechblog.com/per-title-encode-optimization-7e99442b62a2

thank ：

This paper has been approved by the author Krishna Rao Vijayanagar Authorized translation and release , Hereby thank .

Link to the original text ：

https://ottverse.com/what-is-per-title-encoding/

---