Small Pixels Small yet beautiful


Authors: Tiberio Uricchio, Leonardo Galteri, Lorenzo Seidenari, Marco Bertini, Alberto Del Bimbo

Small Pixels stems from research expertise in improving the quality of compressed videos that has been developed over time by the team of proposing members and from trials that have produced encouraging results. Having verified through participation in conferences and events, both the real need of the market for effective solutions and the limited number of valid solutions currently available, the team felt that they could initiate an academic spinoff that after a start-up phase, aimed at the realization of a competitive product, is able to position itself on the international market.

The solution that we want to bring to market is an algorithm to improve the visual quality of images that is implemented in the software used to display videos, such as video players of streaming services and video conferencing systems. Our system allows to view in high quality even videos transmitted in a very compressed format, improving both the user experience and reducing the costs of video management and transmission.

Our solution therefore aims to reduce the cost of video broadcasting, allowing to use more compressed videos without a reduction in the quality of experience and service for the end user of the videos. This can be done by reducing or deleting compression artifacts once the video has been received and decompressed on the device used by the end user at the time of viewing it. The benefit of this approach is that there is no need to change any component of the video processing and distribution chain. In addition, this approach is independent of video compression (codecs) and therefore can also be used with future systems not yet on the market.

The main customer segments that Small Pixels targets are 4:

  • Video Streaming Companies: All companies that stream videos of content, especially for
    entertainment purposes, such as Netflix, TIMVision. These firms need to broadcast video on demand (a copy of the video to each person viewing it) instantly and must adapt to the available bandwidth (ADSL, Fiber, 4G, etc.). Each broadcast of a video has a cost that varies depending on the size of the video and the number of people who view it (even millions of Euros per single episode of a TV serial). Our solution allows a compression of at least 8x, thus potentially reducing transmission costs by 8 times. In addition, when only low-bandwidth is available, videos can only be streamed at low quality with current technologies. With our solution you can have better quality and increase the degree of user satisfaction.
  • Video Infrastructure Providers: All those who store, optimize, and transmit video for third parties, such as Bitmovin, Akamai. Similar to video streaming companies, they resell to third parties the management of large quantities of video, which must be stored, optimized and transferred.
  • Video conferencing: All those who sell video conferencing solutions and who need to broadcast videos of people who want to communicate with each other in real time. When the available bandwidth is low, the videos can be streamed at a very low quality making the user experience disappointing. The proposed solution allows to lower the amount of bandwidth needed to use the service at a good quality and therefore improves the user experience.
  • Video restoring (early adopters): all businesses or private individuals who need to improve the quality of videos they have, such as video producers who need to edit high-quality sequences for use in short/feature films or TV productions but have only low-quality material available. We define these subjects as early adopters as they are also consumers they are easily contactable through advertisements as ad on Facebook, have not too high a number of videos and are inclined to try new technical solutions also through apps, plugins for video editing software or websites.

The main highlights of the proposed solution are:

  • The use of artificial intelligence-based technology for image reconstruction. Existing business solutions use manually engineered reconstruction algorithms that do not consider the content of videos. The proposed algorithm is instead the result of a machine learning process based on the analysis of the content of the videos.
  • The algorithm can be customized based on the video to be reconstructed and the technical needs of the customer. Single video customization is limited in current techniques and would require the development of a new compressor. The proposed solution, instead, being the result of an automatic learning procedure, is able to customize the algorithm on each specific video.
  • High compression rate. The proposed solution is capable of achieving significantly higher compressions (at least 8 times) than current technologies. This results in less space occupancy and lower bandwidth usage and therefore lower costs.

The team is made up of professors and researchers from the University of Florence who have been working together for many years on topics of intelligence and artificial vision with international recognition. This includes the Best Demo Award of ACM Multimedia 2019, the most important international multimedia conference, held in Nice in October, where a live demonstration of the proposed system has won this important award.

More Info HERE

There are no related publications

There are no related projects