This thesis investigates camera calibration methods in a real-time embedded context. The main goal is to propose an innovative processing architecture that enables execution of online autocalibration algorithms in single or multi-views. This architecture will be compliant with embedded system constraints (small footprint and low power consumption) and powerful enough to cope with the execution of applications with high reactivity requirements. In image processing domain, calibration consists in establishing a relation between spatial coordinates of one point in the scene and its equivalent location in an image-based representation captured by one or several cameras. Calibration methodologies (performed either offline or online, with or without a calibration pattern) may differ in complexity or robustness depending on operating conditions. The goal of this PhD thesis is to propose a calibration methodology for image stream under embedded low-latency systems constraints. First step will consist in identifying relevant algorithms that fit embedded systems and applications requirements. Adaptation or improvement of algorithms may be necessary at this stage. A matching step between algorithm and architecture will allow both to highlight hardware/software barriers and to propose innovative answers to these challenges. In particular, this study will focus on memory footprint compaction and latency/accuracy/complexity tradeoffs as well as the scaling up issues (high resolution/framerate, e.g. 4K at 60 fps). This work is also likely to target new opportunities dealing with cutting-edge application such as AR. Required Skills: Embedded Systems, Image Processing, Computer Vision, Architecture-Algorithm Design, Augmented Reality. Key planned tasks: Bibliography (20%), Analysis of existing methods and A3 (30%), Design, simulation and test (35%), Thesis writing (15%).

This position is open until it is filled.

Département: Département Architectures Conception et Logiciels Embarqués (LIST-LETI)
Laboratory: Laboratoire Calcul Embarqué
Start Date: 01-09-2015
ECA Code: SL-DRT-15-0386
Contact: mehdi.darouich<στο>cea.fr