Thesis, internship, and post-doc opportunities

[Thesis]

Smart image sensor architectures, exploiting emergent paradigms of computational optics, for novel acquisition of images

Offer N°: 34519

Start date: 1 Oct 2014

 

This thesis will focus on the co-design of integrated architectures and algorithms for smart image sensors aiming to exploit innovative optical concepts such as open aperture or wavefront coding.

With the development of integrated circuit technologies and advanced signal processing, the concept of smart image sensor has emerged, an area where Leti has developed a great experience. In addition, micro-and nanotechnology are used to make low-cost, while highly precise optics that perform new functions: micro- mirrors, microlens arrays (e.g. plenoptic cameras such as in the Lytro®) or the spatial light modulators (SLM). The exploitation of resulting images then requires complex processing. Currently, conventional techniques of acquisition and processing images fell in exploiting the potential of these optical elements. Indeed, on the one hand imagers are optimized for the acquisition and transfer of conventional images, on the other hand the processing and control of the optics are both performed on remote processors whose architectures cannot process the signals in extremely short and / or time with an economy of means. Relying on the expertise of optical lab partners, the thesis will jointly co-design algorithms and architectures associated with innovative features to overcome state of the art realizations, both in terms of cost, time response or application fields.

Profile: engineer or master 2 with excellent academic records, a strong scientific background and a specialization either in electronic engineering or applied physics, a taste for an interdisciplinary topic at the crossroads between optics, signal processing and integrated electronics.

 

• Laboratory:LETI / DACLE
• CEA code:SL-DRT-14-0396
• Contact: antoine.dupret@cea.fr

[Thesis]

Modeling and dimensioning of hybrid dataflow and real time systems

Offer N°: 34541

Start date: 1 Oct 2014

 

The emergence of new applications mixinf both real-time tasks and computationally intensive “best effort” tasks such as the ADAS applications (Advanced Driver Assistance Systems) or augmented-reality (e.g. Google Glass), has created new research issues in both modelling and execution levels. Indeed, an effective implementation of such applications requires the hybridization of formalisms and techniques for hard real-time with other more suitable to computationally intensive systems. Properties such as determinism, lack of deadlock etc. are not guaranteed in the same way in real-time computing systems and computationally intensive based systems.

The first aim of the thesis is to propose a new task model for these data and time-driven applications. This model will ensure expressing the real-time constraints, guaranteeing high level properties such as determinism, lack of deadlock, liveness, etc. and meeting the size of the buffers.

The second goal is to establish a design flow of mixed dataflow and real-time systems. The task model integration in the design flow includes the robust and timed design of buffers, the characterization of latency induced by a dataflow graph for integration into a comprehensive real-time analysis, real-time proof design of mixed applications, placement/routing with real-time constraints, the automatic generation of embedded code and possible high-level optimizations.

 

• Laboratory:LETI / DACLE
• CEA code:SL-DRT-14-0312
• Contact: kods.trabelsi@cea.fr

Source:

http://www.minatec.org/en/offres?page=11