ESR 14 – Emanuele Palma

In this project, together with my supervisor, we studied a new immersive media: light fields. These images can be represented by a so-called plenoptic function, which allows the viewer (through processing) to render the scene from multiple points of view (angular views). These angular views, which show the same image from slightly various angles or camera placements, can carry a lot of extra information, especially in safety critical environments. 

Using well designed geometry-aware approaches, the microlens structure of a light field capturing decive can be exploited to do very precise image compression, keeping the compressed image with the same quality as the acquired image, enabling the creation of high-performance coding methods that can store losslessly any light field image.

We divided the project in two macro-sections, following the milestones: Efficient Compression Methods and Quality Analysis and Assessment.

In the first section we studied new methods for lossless coding of light fields using sparse linear predictive models and context adaptive arithmetic coding. We also presented a new algorithm for near-lossless coding of light field sensor images.

In the second section we presented a quality analysis and assessment experiment, that shed light on how end users perceive the compression artifacts when it comes to novel imaging structures, such as light fields. A standardized experiment was carried out and binary preference scores were collected to achieve a coherent and informative outcome, that could help us identify the problems in the fruition of such content.