Andrei-Cristian BARBOS, PhD, Image Quality Expert at DXOMARK
Andrei, Thanks for the interview
Thèse (publiée sur theses.fr)
Please describe briefly your thesis subject.
To put it simply, my thesis subject is generating random numbers, or pseudo-random numbers to be more precise. Strange as it might seem, being able to generate random numbers plays an important, even essential role in many applications. For example, security in the digital world (encryption, cryptography) relies heavily on random number generation.
My work was motivated by inverse problems in imaging. An imaging device, be it a smartphone or an MRI, has its physical limitations. An inverse problem in imaging seeks to circumvent the physical limitations of the imaging device and provide an image as close as possible to the ideal one.
An approach to solving such problems is to model the direct process to capture an image as a random process. The inverse problem is consequently stochastic in nature. The set of all possible solutions is expressed as a probability distribution. Solving the inverse problem equates to extracting a point estimate from the distribution.
For problems in imaging, in which each pixel/voxel is considered to be random, the dimension of the probability distribution is very high. Even if we have analytical formulas for the point estimates, it is not feasible to compute them directly. My work proposed a numerical sampling algorithm for Gaussian distributions that is efficient both in computational complexity and also storage complexity in high dimensional settings.
Could you please describe your background?
My journey all started when I got my first computer. This was before high-school. At the time, I didn’t understand how it worked, but I was fascinated and wanted to learn more. When it was time to go to high-school, it was not difficult for me to select a class with information science as specialisation. My under-graduate studies went along the same path. I got an B.Eng. degree in Electronics, Telecommunication and Information Technology from the Technical University of Cluj-Napoca in Romania.
By the end of the under-graduate studies I had a pretty good understanding of how computer works. I was still curious though to learn more about how we can put computers to our use. One field that caught my attention was Signal and Image processing. I enrolled in a master program focused on Signal and Image processing and, thanks to an agreement between the university in Cluj-Napoca and the university in Bordeaux, I had the possibility of an exchange year at the University of Bordeaux. I obtained a double diploma.
It was during my exchange year in Bordeaux that I was more and more attracted into research. I did a research internship at IMS Laboratory in Talence and at the end of the internship my tutor proposed to me to continue with a PhD.
Doing a PhD is not an easy undertaking. What is your impression of your time as a PhD student?
I agree that doing a PhD is a difficult undertaking. On a scientific level, you are pushing the boundaries of your field of study. On a personal level, you are pushing your boundaries and finding your limits.
Overall, it was a fulfilling experience. On a scientific level, we managed to propose a new algorithm for efficient sampling in high dimensional settings. The journey had its ups and downs, though all is forgotten when you reach your goal. On a personal level, I managed to make new friends, from all over the world (Europe, Africa, Asia, Americas), I even managed to meet people from my own home university in Romania that I didn’t have the chance to meet during my under-graduate studies. I also fulfilled teaching duties during my PhD studies period. Teaching is something that I’m very fond of and I was happy that I could share my knowledges and have a positive impact on the academic and personal development of my students. Being a teacher is not only about sharing knowledge, but also about forming individuals.
One particular aspect that I realised during my PhD studies, and that many might find strange, is that the more you get to know, the more you realise that overall you don’t know much. This is even more visible when you attend scientific conferences and get to discuss with world renowned scientists. You feel humble when discussing with such individuals, but you also get motivated to further your game and hope that one day you will be among them.
What is you current position and how does the PhD provide useful in you daily job?
I currently occupy the position of Image Quality Expert at DXOMARK Image Labs. The main activity of the company is to provide consulting services to giant tech companies in improving the image quality level provided by their devices (smartphones, laptops, etc.).
I don’t necessarily have a daily usage for my expertise in sampling algorithms. However, a PhD is more than just acquiring very specific knowledges about how to perform a certain task, i.e. generate random numbers in my case. The set of skills that you develop during your PhD is large, both technical and soft skills. Some of the skills that I developed during my PhD and that I use on a daily basis are given below:
· deep understanding of the image acquisition chain with the inherent limitations and the solutions to overcome them
· scientific rigour and attention to details when designing and executing the test cases for the devices that we test
· Efficient communication and presentation skills of the test results
What is your advice to a prospective PhD student?
I know it can be daunting to engage in a PhD program, I also had my share of doubts before starting. Nothing comes easy in life, more so finishing a PhD, but the rewards are more than worth the effort. A PhD is an investment in oneself, an investment that usually pays hefty dividends later on in ones career.