"It seems that the old man came"
"Spring should be good, if you are still present"
"Never seen in my life, I will miss it for a lifetime"
Last night, "Zhang Guorong's 2000 "Hot Love" Concert Super Qing Restoration Edition" premiered worldwide on platforms such as Video Number, QQ Music, Kuwo Music, Kugou Music, National K Song, Tencent Video, and Aurora TV.
This 21-year-old master tape comes from PolyGram, a label owned by Universal Music.
Limited by the storage medium of shooting equipment and video materials at that time, there is room for improvement in pictures and sounds.
This time, Tencent uses a number of industry-leading self-developed technologies such as the ultra-clear immersive restoration engine to present you with a better sound and picture live broadcast experience.
Before and after the repair
Today, we will talk about:
The video from 21 years ago is how to carry out ultra-clean restoration.
The secret to repairing the old as old and getting a sense of the scene
The first step is "intelligent analysis", which is completed by artificial intelligence.
It's the equivalent of doing a full scan.
The whole video is close to two hours, and there are many scenes.
In order to make the analysis results as complete and accurate as possible, we have set up multiple algorithms, covering multiple dimensions such as frame level, scene level and video level.
After some analysis, the "scan report" was released:
Where there is noise, where the color is wrong, where the compression distortion is...
Then, set goals -
Repair the old as old.
The purpose is to bring fans a "live" viewing experience.
For such repair scenarios, Tencent Cloud Multimedia Lab has launched a technical solution:
Ultra-clear immersive restoration engine.
This system is very clever -
It can identify which are the "hard wounds" that need to be repaired, and which are the "immersion" and "nostalgia" that need to be retained.
How to bring the picture back to HD
It's time for the formal picture restoration.
Using intelligent algorithms, we have fixed a variety of factors that disrupt quality content.
For example, problems such as vertical lines, snowflake spots, noise, compression distortion, and jitter that are common in old films.
The results are obvious.
Deinterlace before and after comparison
Video goes from shooting to storage; from film to video tape, to digital film... Various processes face "compression distortion".
So, one of the steps to fix it is:
Return the video to its pre-compression state, which is "de-compression distortion".
Through the algorithm model, various distortion artifacts caused by video encoding, video processing and other reasons are learned, so as to remove the distortion type of recoded video, restore picture details and repair edge loss.
(A bit of a mouthful, to see the effect)
Remove compression distortion before and after comparison
Faces reproduce real detail
After using face restoration techniques, facial details become clearly visible.
Zooming in on the picture, you can see the outline of the image refracted on the stage through your brother's eyes.
It's still.
Video, on the other hand, is made up of many pictures.
Therefore, it is a problem to keep the repair effect continuous and stable.
In response to this problem, we have developed our own face repair model -
Using the Face Detection, Face Repair and Fusion modules, image repair is migrated to the video process and inter-frame stability processing is added.
This technology solves the stability of face detection and optimizes the multi-angle face effect.
Finally, we will also improve the multi-dimensional picture quality of the picture -
Boost video resolution and frame rate in space and time, and improve detail clarity and color richness in airspace.
This link, called "enhancement", is also an easy step.
And one of the tricks of "fixing the old as old" is also hidden in this step -
The degree of color enhancement should not be too large.
Enhance front-to-back contrast
True idols, will not leave.
How much time has passed, still sound and smile.
Images are the medium through which good memories are stored.
With the help of advanced restoration technology, we can make more classic images clearly reproduced.