Tweets by AnaK_DA I am a Research fellow at VSENSE and the Graphics Vision and Visualisation group (GV2) at Trinity College Dublin (TCD), working with Aljosa Smolic. Currently, I am interested in problems related to image/video perception, visual attention modeling and prediction, Omnidirectional video and VR.
Previously, I was a PhD student at LTS4 in EPFL working on interactive multiview video (IMV), in particular on optimizing coding strategies for IMV. In EPFL, my work was supervised by Pascal Frossard and Fernando Pereira. Before my PhD, I did an intership at Cisco, where I worked on video quality assessment.
You could also check my Google scholar account.
In order to predict salient image regions, saliency maps are estimated either by using an eye tracker to collect eye ﬁxations during subjective tests or by using computational models of visual attention. In this work, we collected viewport data of 32 participants for 21 omdirectional images (ODIs) and propose a method to transform the gathered data into saliency maps. The obtained saliency maps are compared in terms of image exposition time used to display each ODI in the subjective tests. Available demo.
The search for the “optimal” Lagrange multiplier that results in a distortion-minimizing solution among all Lagrangian solutions that satisfy the original rate constraint, remains an open problem in the general setting. To address this issue, we are the first in the literature to construct a computation-efficient search strategy to identify this optimal multiplier numerically in the general dependent coding scenario.
We propose a novel adaptive IMVS solution based on a layered multiview representation where camera views are organized into layered subsets to match the different clients constraints. We formulate an optimization problem for the joint selection of the views subsets and their encoding rates. Then, we propose an optimal and a reduced computational complexity greedy algorithms, both based on dynamic-programming.
Interactive multiview video (IMV) applications offer to users the freedom of selecting their preferred viewpoint, by using intermediate virtual views. However, the virtual views quality depends on the distance to the available views used as references and on their quality, which is generally constrained by the heterogeneous capabilities of the users.This work proposes an IMV scalable system, where views are optimally organized in layers, each one offering an incremental improvement in the interactive navigation quality. We propose a distortion model for the rendered virtual views and an algorithm that selects the optimal views' subset per layer.
This paper proposes an algorithm for the effective selection of the interview prediction structures (PSs) and associated texture and depth quantization parameters (QPs) for interactive multiview video streaming (IMVS) systems under relevant constraints.
Multiview Video Coding (MVC) has been developed to efficiently compress a set of camera views. However, the resulting compressed data has a lot of prediction coding dependencies, which may not suit interactive multiview video streaming (IMVS) systems. This paper proposes a fast selection mechanism for effective interview prediction structure (PS) in IMVS while minimizing the point-to-point transmission rate, given some storage and visual distortion con- straints, and a user interactive behavior model.
This letter studies the impact of the MVC interview prediction structure on both the transmission and the overall coding rates for an interactive multiview video streaming system, considering both unicast and multicast scenarios, with the user interactive behavior represented by some view-popularity model.
Working with Aljosa Smolic on perception issues on VR.Jul, 2016 - Now
Video streaming quality assessment.May, 2010 - Mar, 2011