Subject title: Image denoising of FIB/SEM images mixing inverse problems and deep learning based methods, with application to failure detection and analysis in microelectronic devices.
Host laboratory: Laboratoire Hubert Curien (LaHC), 18 Rue Pr B. Lauras, 42000 SAINT-ETIENNE.
Supervisor: Olivier Alata (email@example.com), Fabien Momey (firstname.lastname@example.org).
Keywords: Image denoising, inverse problems, deep learning, scanning electron microscopy (SEM), failure detection analysis, microelectronic devices, european project.
Duration: 18 months.
Starting date: January 2021.
Salary: ~ 2200 euros/month net (for experience post-PhD <= 3 years). Teaching activities are eventually possible
Application deadline date: October 19th 2020
Decision announcement date: October 23th 2020. Nevertheless, applications will be analysed upon receipt.
Your application should include the following documents:
• Letter of intent
• Scientific CV
• List of publications
• Names of Referees (at least 2)
• Olivier Alata <email@example.com> (web page)
• Fabien Momey <firstname.lastname@example.org>
Context and problematics:
High-tech products based on electronic devices benefit from rapid progress in computer science and automation, leading to more complex embedded systems (autonomous driving, mobile telephony, connected objects, etc.). This has given raise to the need of extreme reliability of electronic components, for daily uses. Being able to efficiently detect, characterize and analyze failure and defects is then mandatory to ensure this reliability all along the lifetime of the component, from the conception and fabrication to the integration and use in embedded systems.
The global objective FA4.0 (Failure Analysis 4.0) european project is to benefit from recent advances in algoirthms of artificial intelligence (AI) (machine learning, deep learning) to develop a complete pipeline for failure diagnostic of electronic devices, including the constitution of an "intelligent" data warehouse and all their associate processing (data mining), that could allow automated recognition and analysis of defects.
The consortium includes 4 european countries (Germany, France, Czech republic, Sweden) that involves 3 leader companies of the electronic systems industry (Infineon, ST Microelectronics, Bosch), 7 medium-sized societies (imaging devices, computer science, ...) and 3 research laboratories (including Laboratoire Hubert Curien of Saint-Etienne University).
Objectives of the Post-Doc: image denoising of FIB/SEM images
This Post-Doc makes part of one of the tasks of the workpackage "Hardware and AI algorithms for advanced signal processing and defect identification", that concerns image denoising for various sensors. Specifically, the task involves a collaboration between ST Microelectronics, the Laboratoire Hubert Curien and the company Orsay Physics (Tescan). This last is specialized in the conception of scanning electron microscopy (SEM) devices, coupled with a particular technology called focused ion beam (FIB). This is a technique of choice for failure detection, which uses a high-energy ion beam to perform matter ablation on a chip (sample slicing for inspection), while simultaneous imaging of the sample at a nanoscale is performed thanks to a SEM column (https://en.wikipedia.org/wiki/Focused_ion_beam). SEM/FIB can also be used for real-time controlled circuit editing.
A correlated work with colleagues at Laboratoire Hubert Curien has been initiated by another recruited post-doc researcher to study the physical process of the SEM detector chain (collection of secondary electrons -> photon conversion with a scintillator -> amplification by photomultiplication -> conversion in the recorded electrical signal), and propose a noise modeling, that can be considered, in first approximation, as a cascade of Poisson processes.
The objective is to benefit from this work to elaborate physically reliable denoising methods for SEM images. To this aim, the recruited candidate will first review the state-of-the-art in image denoising techniques applied to SEM, and will explore several "paradigms" such filtering , inverse problems approaches [2,3], and recently-appeared deep learning based methods [4,5,6]. The originality of this work could rely in finding a mixed approach jointly based on inverse problems and deep learning approaches.
The post-doc will work in The Hubert Curien Laboratory Image Science and Computer Vision team, which forms a chain of skills in image processing and analysis, gathering the physical and semantic approaches. The team investigates main links of the image chain:
- image attributes and reproduction (modeling and metrology of optical surfaces),
- imaging systems (image formation, restoration and reconstruction),
- image interpretation and understanding (multidimensional analysis and machine learning).
Required skills: signal and image processing, image denoising, inverse problems, deep learning. A background in image processing dedicated to SEM or similar imaging techniques could be of interest.
 Kostadin Dabov, Alessandro Foi, Vladimir Katkovnik, and Karen Egiazarian. Image Denoising by Sparse 3-D Transform-Domain Collaborative Filtering. IEEE Transactions on Image Processing, 16(8):2080-2095, August 2007. Conference Name: IEEE Transactions on Image Processing.
 A. Beck and M. Teboulle. Fast Gradient-Based Algorithms for Constrained Total Variation Image Denoising and Deblurring Problems. IEEE Transactions on Image Processing, 18(11):2419-2434, November 2009.
 Mujibur Rahman Chowdhury, Jing Qin, and Yifei Lou. Non-blind and Blind Deconvolution Under Poisson Noise Using Fractional-Order Total Variation. Journal of Mathematical Imaging and Vision, 62(9):1238-1255, November 2020.
 Kai Zhang, Wangmeng Zuo, Shuhang Gu, and Lei Zhang. Learning Deep CNN Denoiser Prior for Image Restoration. pages 3929-3938, 2017.
 E. Giannatou, G. Papavieros, V. Constantoudis, H. Papageorgiou, and E. Gogolides. Deep learning denoising of SEM images towards noise-reduced LER measurements. Microelectronic Engineering, 216:111051, August 2019.
 Christopher A. Metzler, Ali Mousavi, Reinhard Heckel, and Richard G. Baraniuk. Unsupervised Learning with Stein's Unbiased Risk Estimator. arXiv:1805.10531 [cs, stat], July 2020. arXiv: 1805.10531.
(c) GdR 720 ISIS - CNRS - 2011-2020.