Ismael Daribo | Scientific Researcher

In the framework of 3D multiview imaging system, a large number of cameras capture the same scene from different viewpoints. There exists a fairly strong correlation between the different video sequences, which leads to the necessity to properly take it into account, specially if one wants to transmit such 3D video sequence on a communication channel throughout the different views. Current 3D video standards use conventional predictive patterns, where the prediction is no longer calculated on the time axis but also from one camera to another one.

This thesis aims to study the technical processing of a 3D video communication system, including both broadcast model (coding for video 2D + depth) and multiview video model. Direct applications are 3DTV and FVV applications. This area of research gathers various known techniques as video coding 2D+t theory (e.g., predictive coders, wavelet coders, etc.), computer vision background related to multiview imaging (e.g., stereoscopic analysis, DIBR, etc.), 3D display technologies, etc.

download the dissertation.

@PHDTHESIS{Daribo_09,
  author = {Isma\"{e}l Daribo},
  title = {Coding and rendering of 3D video sequences;
  and applications to Three-Dimensional Television (3DTV)
  and Free Viewpoint Television (FTV).
  },
  school = {Graduate College of Telecom ParisTech},
  year = {2009},
  address = {Paris, France},
  month = nov,
  keywords = {3D video communication, three-dimensional television, 3DTV,
	free viewpoint television, FTV, three-dimensional phone, 3D phone,
	MPEG-2, MPEG-4 AVC, H.264, MVC, video-plus-depth, texture-plus-depth
	multiview video, MVV, free viewpoint video, FVV video coding, image coding,
	texture coding, depth coding, depth compression, 3D video coding,
	multiview video-plus-depth, texture-plus-depth coding, motion estimation,
	joint motion estimation, rate allocation, joint rate allocation,
	distortion model, view synthesis, view rendering, rendering (computer
	graphics), multiview video-plus-depth virtual view rendering, virtual
	intermediate camera view, multiview rendering quality},
}