Image / Video Compression

Nowadays we are assisting to unprecedent time in history, where we are surrounded by a large number of digital devices helping us to deliver in a much more efficient way than before our daily activities, i.e., working, recreation, personal etc. Many of these activities has to do with the streaming of large data composed by images and videos, and at the same time the recreation industry is continuously increasing the level of experience to convey a better user experience, i.e., HD vs. 4K UHD vs. 8K UHD, SDR vs High dynamic Range (HDR). This has been even increased during the pandemic phase that we are facing right today. To cope these needs, standard organizations are spending considerable time in developing new encoding standards technology, which still is not able to satisfy the overall needs requested by the market/applications. Maximizing video quality, will require to compromise the compression efficiency. Maximizing the compression efficiency, will require to compromise the video quality. So, a good tradeoff between quality and compression efficiency needs to be found among several parameters, e.g., bandwidth, costs to deliver the service etc. On the other hand, deeplearning approaches are proposed either to substitute blocks, i.e., intradcoding, downsampling/upsampling blocks,  into an existing compression standards, i.e., HEVC, or as end-to-end approach, as mean of providing a substantial improvement of the overall quality of the compression standard itself.

AI Based Media Coding

Our group as founding member of the MPAI standard is participating in the exploration phase for investigating whether it is possible to improve the performance of the Essential Video Coding (EVC) modified by enhancing/replacing existing video coding tools with AI tools keeping complexity increase to an acceptable level.

On Going Activities

In order to achieve our goals to enhance the traditional video coding tools with AI, out team is currently working on providing a solution for Super Resolution Problem that is comparable to state of art models with lesser memory consumption and better time efficiency

Super Resolution

Super Resolution is the process of recovering a High Resolution (HR) image from a given Low Resolution (LR) image. It is an important class of image processing techniques in computer vision and image processing and enjoys a wide range of real-world applications, such as medical imaging, satellite imaging, surveillance and security, astronomical imaging, amongst others. With the advancement in deep learning techniques in recent years, deep learning-based SR models have been actively explored and often achieve state-of-the-art performance on various benchmarks of SR. A variety of deep learning methods have been applied to solve SR tasks, ranging from the early Convolutional Neural Networks (CNN) based method to recent cascading residual architecture.

Our team is working on simplification of recent state of art model i.e. Densely residual Laplacian Network. it has been found through our rigorous testing that a simplified version of proposed model not only reduces memory consumption and increase time efficiency but also provides comparable results.

Thus, the goal of super resolution can be achieved with less complex deep learning architecture.