High-definition Color Film of the Founding Ceremony of the People's Republic of China Was Released: With the Support of Technology, History Will Not Fade

In the movie "Call of Duty" which will be released tomorrow, high-definition historical materials that have been restored in 4K will be available to the people across the country. This hard-won material was obtained by the creators after a lot of restoration work. What exactly is film restoration, and how can AI help in this process?

This year marks the 70th anniversary of the founding of our motherland. If there is a video that can show usHigh-definition color opening ceremony,It is sure to bring tears to the eyes of countless people.

This is not just a random thought.High-definition historical materials restored in 4K,He will appear in the movie "Calling Moment" which will be released tomorrow.

Let’s take a look at the actual effect after the repair:

According to the film producers, this material comes from a Russian color documentary, but due to its age,The picture quality is blurry and the colors are distorted.

After complex 4K restoration work, the final result is such an amazing effect, allowing us today to have the opportunity to witness the founding ceremony of the country 70 years ago.

Technology that helps us see classics again: Film restoration

The reason why we can see such clear historical images is due to a process called Film Restorationtechnology.

Restoration is also called remake, which usually refers toUsing technical means, old movies such as film can be transferred to new carriers to achieve higher clarity and viewing experience.

By usingRepair, noise reduction, fill light, color gradingThe names of 4K and 2K refer to the restoration standards based on the resolution.

In the more than 100 years since the advent of movies, their production and shooting techniques have been constantly improving. For example, the material of early films has been updated, silent films have been transformed into sound films, black and white has been transformed into color films, the emergence of digital films, and now IMAX-3D. These technological changes have all driven the development of movies towards better light and shadow presentation.

As film movies exit the stage, early film movies in history are also being lost by time. Due to technical defects, lack of preservation awareness and other reasons, some of them have been lost, while others have been damaged and the image quality no longer meets the requirements for screening.

In order to ensure that classic old movies will not be forgotten and to bring them back to life on the screen,The technical fix became the final straw.

In 2006, the country launched "Film Archive Digital Restoration Project", investing a lot of money to promote China 40,000Digital restoration and archiving of old film films.

But film restoration is a complex project.The typical repair time is from a few weeks to a few months, and sometimes even up to two years.At this rate of restoration, some film reels may not live to see the day of restoration.

Therefore, new restoration techniques and technologies have emerged, and AI has naturally joined the ranks to contribute to film restoration.

Film restoration is difficult, and human labor cannot outrun time

Through restoration technology, many old movies and video materials are presented in a clearer form, such as "Gone with the Wind", "A Chinese Odyssey", and even the animated film "My Neighbor Totoro", which have caused a great response.

But in fact, behind the film restoration is a huge project.Physical repair,Remove impurities from the film and thenDigitalization of film materials,ExpandDigital restoration,Restoring movie images frame by frame.

There are six main types of damage to film: dirt, scratches, noise, mildew, jitter, and flicker. Film restoration is actually to eliminate and correct problems on the film, and pay special attention to restoring the original texture of the film.

The latest example of film restoration is the National Day film The Founding Ceremony,Its 4K restoration is already scheduled.

This classic film was released in 1989. 30 years later, the original high-quality production can no longer keep up with today's image quality requirements.

After 4K restoration, the effect of "The Founding Ceremony" is gratifying, but director Li Qiankuan lamented the difficulty of film restoration. He said: "The workload of this restoration is huge, and the restoration team needs to  240,000 framesThe images were restored frame by frame.

And the biggest challenge of restoration comes from the film 72Historical footage, although onlyTen minutes,But poured 600 peopleThe whole restoration team spent a lot of effort 70%  time.

It can be seen that repairing it purely by manpower is a difficult process. Fortunately, with the development of technology,AI has brought about qualitative and quantitative improvements in film restoration.

Abroad: Deepsense AI restores "Gone with the Wind"

A family called Deepsense AI The company recently used AI to restore the classic film "Gone with the Wind".Convolutional Neural Networks and Image Recognition,Bringing back the glory of such old movies.

Classic movies are usually shot in low resolution, and over time, the original film will become aged, full of noise and cracks.Eliminate distortionThe Deepsense AI team usedMachine Learning Algorithms, which solved the problem of automatically restoring film to high-definition digital format. They also adoptedSupervised Learning Methods, removing cracks and noise from digitized versions of movies.

The principle of the method is to use deep neural networks to reduce the resolution of high-definition movies, conduct supervised training on the model, and let AI learn to present blurred details.

In addition, to train the model, the researchers manually added cracks and defects to the film, and combined with images of the original defective film, the model learned how to remove those defects.

The neural network can also process and repair severely damaged original film. When the image in the picture is too dark and blurry to be recognizable, the network can generate low-quality predictions.
These methods have successfully completed the high-definition restoration of old movie images, so that classic old movies can be presented clearly to the audience. 

Domestic: "Shensi" system repair "Great, My Country"

in the country, there are also teams dedicated to using AI for film restoration.

For example, the China Film Digital Production Base developed "Sensing" AI image processing technology,The film "Great, My Country" was restored to create a high-quality documentary.

The material for this documentary is taken from the earlier six-episode documentary "Glorious China", so the video was originally made for television, with low resolution and poor picture quality.

In order to be shown in theaters, the small screen with a resolution of 720×576 needs to be turned into a large screen with a resolution of 2K or even higher in the theater, and the shortcomings and problems in the picture will be magnified.

To solve this problem,University of Science and Technology of ChinaThe technical team cooperated with the China Film Base to build "Shensi".After being trained with hundreds of thousands of videos and millions of images, AI has learned to repair videos autonomously.

They useDeep learning algorithms,Using adversarial generative networks GAN, DenseNet, CycleGAN  This system was built using network and other technologies.

The entire framework consists of three modules: resolution enhancement and image quality enhancement processing unit, standard definition image de-fielding processing unit, and image digital restoration processing unit.

The highlight of the system is that it has conducted a large amount of material training for different image processing problems, used neural networks to maximize image quality, automatically processed large quantities of image materials, and created a recurrent network to improve the system's computing efficiency and stability.

Through the application of these technologies, the Shensi system can processTens of thousands of frames, even millions of framesIn the end, they processed more than 10,000 images of "Great, My Country" in four months.Three hundred thousand framesLow-resolution image material.

Expect AI to make more old movies shine

In film restoration, the application of AI technology not only allows us to see old movies in high definition again, but also greatly improves efficiency and liberates manpower. The originally heavy labor has become a nearly "one-click" operation.

In addition to the image repairs listed in the article,AI has also achieved good results in movie sound restoration, black-and-white movie colorization, and picture-sound synchronization.Helping old movies come back to life.

But on the other hand, film restoration cannot rely entirely on technology. AI currently only repairs and enhances images or sounds, but movies also haveThe elements of art,The extent of the repair still needs to be determined manually, and the algorithm cannot grasp the right measure for the time being.

In the industry, the highest level of film restoration is "Restore the old to its original state"How to preserve the original flavor of old movies while improving image quality may be a direction that AI still needs to work on. 

-- over--