Research Projects

The current decade is witnessing a meteoric rise in creation and usage of new kinds of multimedia technologies. Media formats such as light fields, 360° media, point clouds, etc. are new and are being thoroughly explored both by the research community and industries as well. These technologies are inherently different from the traditional 2D imaging. For instance light fields allow the viewer to freely focus on any plane which the viewer wants to explore. Similarly, point cloud data projects millions of points onto a single surface, thus generating a 3D object model. The 360° media allows telepresence of a viewer to a virtually remote environment with an immersive feeling to the viewer. It can be noted here that there is a common factor which connects all the three above mentioned new kinds of multimedia technologies. These new media formats target to provide real-life experience to the viewers of the multimedia content and volumetric data points for accurate processing. There are multiple application areas where this new media is frequently used. Starting from VR platforms for gaming, education, medical purposes, etc. In various hazardous safety critical applications, VR platform can simulate the environment and thus enabling a new trainee to learn the critical equipment without any risk. Since these media formats are typically delivering real-life experiences, the quality of the generated content is of extreme importance. Perceptual quality or the Quality of Experience (QoE) of a multimedia content is the measure of delight or annoyance of a viewer's experience. It tries to estimate which signals in a media detracts the enjoyment or fulfillment of a visual experience. In this regard, recent literature has studied specific properties of 360° and proposed a few quality metrics. However, the metrics are evaluated on data that are generated in a Laboratory setting or typical computer generated animated environment. Incidentally, there are multiple reasons due to which the existing metrics evaluated over guided media content may fail in real-world application, There exist limited quality metrics for 360° and they are not suitable for random user generated 360° content (RUG360C). Nowadays handheld devices are available at a very low cost. Thus such sophisticated devices are now frequently used to capture 360° content for sharing in public platforms or storing for personal usage. These are consumer-generated content and hence are severely degraded with artifacts and reference images are not available for assessment. Such media multiple distortions induced due to camera motion, freezing, blurring, etc. which is completely different from a professional content. Thus, there is a genuine need to estimate the signals (features) of the RUG360C which may detract from the viewing experience and lead towards annoyance.