Previous | 
Home
Section: Bilateral Contracts and Grants with Industry
Bilateral Grants with Industry
CIFRE contract with Orange on Generalized lifting for video compression
Participants : Christine Guillemot, Bihong Huang.
-
Research axis : § 7.3.2 .
This contract with Orange labs. (started in April. 2012) concerns the PhD of Bihong Huang and aims at modelling the redundancy which remains in spatial and temporal prediction residues. The analysis carried out in the first year of the PhD has shown that this redundancy (hence the potential rate saving) is high. In 2013, different methods have been investigated to remove this redundancy, such as generalized lifting and different types of predictors. The generalized lifting is an extension of the lifting scheme of classical wavelet transforms which permits the creation of nonlinear and signal probability density function (pdf) dependent and adaptive transforms. This study is also carried out in collaboration with UPC (Prof. Philippe Salembier) in Barcelona.
CIFRE contract with Technicolor on High Dynamic Range (HDR) video compression
Participants : Mikael Le Pendu, Christine Guillemot.
-
Title : Floating point high dynamic range (HDR) video compression
-
Research axis : § 7.3.4 .
High Dynamic Range (HDR) images contain more intensity levels than traditional image formats, leading to higher volumes of data. HDR images can represent more accurately the range of intensity levels found in real scenes, from direct sunlight to faint starlight. The goal of the thesis is to design a visually lossless compression algorithm for HDR floating-point imaging data. The first year of the thesis has been dedicated to the design of a quantization method converting the floating point data into a reduced bit depth representation, with minimal loss. The method leads to a bit rate saving of compared to the existing Adaptive LogLuv transform.
CIFRE contract with Technicolor on sparse modelling of spatio-temporal scenes
Participants : Martin Alain, Christine Guillemot.
-
Title : Spatio-temporal analysis and characterization of video scenes
-
Research axis : § 7.1.2 .
A first CIFRE contract has concerned the Ph.D of Safa Cherigui from Nov.2009 to Oct.2012, in collaboration with Dominique Thoreau (Technicolor). The objective was to investigate texture and video scene characterization using models based on sparse and data dimensionality reduction techniques, as well as based on epitomes. The objective was then to use these models and methods in different image processing problems focusing in particular on video compression. While, the first PhD thesis has focused on spatial analysis, processing, and prediction of image texture, a second CIFRE contract (PhD thesis of Martin Alain) has started in Oct. 2012 to push further the study by addressing issues of spatio-temporal analysis and epitome construction, with applications to temporal prediction, as well as to other video processing problems such as denoising and super-resolution.
CIFRE contract with Thomson Video Networks (TVN) on Video analysis for HEVC based video coding
Participants : Nicolas Dhollande, Christine Guillemot, Olivier Le Meur.
-
Title : Coding optimization of HEVC by using pre-analysis approaches.
-
Research axis : § 7.3.5 .
This contract with TVN (started in Oct. 2012) concerns the PhD of Nicolas Dhollande and aims at performing a coding mode analysis and developing a pre-analysis software. HEVC standard is a new standard of compression including new tools such as advanced prediction modes. Compared to the previous standard H.264, HEVC's complexity is three to four times higher. The goal of this thesis is to infer the best coding decisions (prediction modes...) in order to reduce the computational complexity of HEVC thanks to a pre-analysis step. The pre-analysis is expected to provide useful estimates of local video characteristics which will then help selecting the prediction and transform partitions as well as a number of other parameters such as the quantization parameters or the prediction modes.
CIFRE contract with Envivio on LDR compatible HDR video coding
Participants : Christine Guillemot, David Gommelet, Aline Roumy.
-
Research axis : § 7.3.3 .
The goal of this Cifre contract is to design solutions for LDR-compatible coding of HDR videos. This involves the study of rate-distortion optimized tone mapping operators taking into account constraints of temporal coherency to avoid the temporal flickering which results from a direct frame-by-frame application of classical tone mapping operators. The goal is also to design a coding architecture which will build upon these operators, integrating coding tools tailored to the statistics of the HDR refinement signals.
CIFRE contract with Technicolor on light fields editing
Participants : Christine Guillemot, Matthieu Hog.
Editing is quite common with classical imaging. Now, if we want light-fields cameras to be in the future as common as traditional cameras, this functionality should also be enabled with light-fields. The goal of the PhD will therefore be to develop methods for light-field editing focusing first on object removal thanks to light-fields inpainting and for constructing panoramic images based on light-fields stitching. This objective also includes the development of algorithms for dynamic light fields spatio-temporal segmentation with spatio-temporal coherence constraints across sub-aperture images.
CIFRE contract with Technicolor on cloud-based video compression
Participants : Jean Begaint, Christine Guillemot.
The goal of this Cifre contract is to develop a novel image compression scheme exploiting similarity between images in a cloud. The objective will therefore be to develop rate-distortion optimized affine or homographic estimation and compensation methods which will allow us to construct prediction schemes and learn adapted bases from most similar images retrieved by image descriptors. One issue to be addressed is the rate-distortion trade-off induced by the need for transmitting image descriptors.