Accueil du site > Home page > Research Departments and Teams > Image Department > → VAADER Team

VAADER (Video Analysis and Architecture Design for Embedded Resources) Team

Toutes les versions de cet article : English , français

Contact persons
Team leader : Luce Morin (PR, INSA)

Team assistants : Corinne Calo (INSA) - Jocelyne Tremier (INSA)

Team : List of members


Presentation of the team

The Video Analysis and Architecture Design for Embedded Resources (VAADER) team studies all aspects of image and video processing, from algorithms conception to their implementations on embedded systems.

The VAADER team focuses on 4 topics :


    - Topic 1 : Image analysis and understanding
    - Topic 2 : Video representation and compression
    - Topic 3 : Rapid prototyping
    - Topic 4 : Data security

 

 

Topic 1 : Image analysis and understanding

- Face detection and tracking (in collaboration with FAST team)

    • Detection, analysis and classification of facial micro-expressions
    • Detection and tracking of objects in a video
    • Detection and analysis of visual saliency


- Embedded vision (with topic 3)
    • Analysis of structured 3 environments : indoor scenes and urban environments.
    • 2D/3D image/model registration for Augmented Reality.
    • Servoing for a semi-autonomous wheel-chair.
    • Implementation of vision algorithms on embedded and parallel architectures.
Semi autonomous wheel-chair
Semantic segmentation

 

Topic 2 : Representation and coding of videos

- Post-HEVC video compression
    • Algorithmics for future video compression standards.
    • Compression for new video formats : 3D video, UHD, HDR, HFR.

- Visual quality
    • Visual quality based video compression.
    • Design of objective metrics for subjective quality.
    • Subjective test campaigns for image and video quality evaluation.

- Video compression on embedded systems (with topic 3)
    • Low–power consumption video compression.
    • Real time implementation of video codecs.

- Video contents security (with topic 4)
    • Image and video data protection methods integrated into compression.
Multi-view plus detph data

 

Topic 3 : Rapid prototyping for parallel and embedded architectures

- Data-flow modeling
    • Design and implementation of data-flow based compilation scheme : PREESM .
    • Runtime support for reconfigurable dataflow : Spider.

- Energy optimization for digital systems
    • Complexity reduction : approximate computing.
    • Optimization of processor clock frequency.

- Implementation of signal and image algorithms (with topics 1 and 2)
    • Data-flow modeling of video applications.
    • Design and implementation of low-power video codecs.
    • Real time implementation of video codec : MPEG4, AVC, SVC, HEVC, SHVC.
    • Experiments on various architectures : Texas Instruments Keystone (DSP), Samsung Exynos, Xilinx Zynq (SoC), Kalray MPPA (Many‐cores).
Data-flow model of a stereovision application using PREESM
Kalray MPPA platform Low-power real-time video decoder

 

Topic 4 : Data security

- Chaos-based approaches
    • Chaos-based generators
    • Chaos-based crypto (block/stream ciphers) and crypto-compression systems.
    • Chao-based steganography and watermarking.
    • Cryptanalysis and universal steganalysis. • Chao-based keyed hash functions.

- Application to security of image and video data
    • Chaos-based crypto and crypto-compression in standard codecs.
    • Robust watermarking.
    • Security of IoT.

International collaborations
The VAADER team established long-term collaborations with researchers from all the continents, through co-directed PhD theses, incoming and outgoing mobility, scientific collaboration and common publications, collaborative software development, partnership within collaborative projects, common actions within standardization, master-level common educational tracks, … Here are some particularly active collaborations :


    - Around MPEG-RVC (Reconfigurable Video Coding), the team collaborates with EPFL Lausanne, Switzerland ; UPM Madrid, Spain ; University of Maryland, USA ; Abo Academy University, Finland and University of Cagliari, Italy ; University of Sassari, Italy.

    - The team collaborates with AUT (Auckland University of Technology) on the thematic of embedded vision and multi-processor architectures.

    - A durable collaboration has been established with University of Sidi Bel Abes, Algeria, image analysis and compression…

Industrial collaborations
The team has strong collaborations with several industrial partners. The team is involved in regional projects within the Breton « Images et Réseaux », which led to increase collaborations with local SMEs such as Nexter Systems, Advancee, within projects supported by the Brittany Region, as well as collaboration with IRT B-com. On the national scale, the team is involved in several collaborative project supported by ANR and DGCIS, with industrial partners such as Orange Labs, Technicolor, Thales, ATEME, EDIXIA, LOGIROAD. At the European scale, the team participates to 3 European projects (4K-REPROSIS, CERBERO, ADAPT). The team has also participated for a long time to bilateral partnership within CIFRE PhD thesis collaboration with large companies (Technicolor, Orange Labs, TDF, STMicroelectronics, Ericsson…).


VAADER (Video Analysis and Architecture Design for Embedded Resources) Team

Scientific production Thesis in progress
 
Technical platforms






Départements et équipes de recherche
Top