This period covers:- Image Sciences Institute- Scale-space research in Utrecht- European Congress of Radiology
These were pioneering days. Not only did we make the first digital reading room in Europe with a working digital Picture Archiving and Communication System (PACS), we were also member of the first medical 3D visualization research groups.
We even figured on the frontpage on 'Diagnostic Imaging', August 1997,a widely-read magazine.
For the Department of Radiology I purchased the Analyze^TM image analysis software from de Mayo Clinic [Robb1989], which was ideal in visually triggering ideas from the radiologists. I also gave many guided tours for students, showing medical image analysis examples with this tool. Some early projects involving image processing, among others, were:
Analyze 14.0 (today)
Rubber sheet matching with Venot's method
Principle of digital subtraction angiography: Make two X-ray images, one without contrast, and the other woth contrast medium in the vessels. Subtraction gives only the vessels.
And then we could make a big step:
In 1984 prof. Stephen Pizer, head of the Medical Image Display & Analysis Group (MIDAG) at the University of North Carolina, US, spent a sabbatical leave with prof Koenderink at Utrecht University [PDF: Nato Grant 1984 enabling multiple mutual visits between UNC and Utrecht]. Steve became a close friend. I could always stay in his home, and he was the best man at my marriage in1988. He donated us the full C++ software environment of his UNC group [Gash1992], [Appelman1993], so we could make a head start with local image analysis support to Radiology. This software was one of the founding packages of the famous ITK software, originally developed for the Visible Human Project.
Prof. Stephen Pizer, Utrecht 1984. [Homepage at UNC]. Now Fellow of the MICCAI Society. He made important discoveries in object representation and segmentation.
The grant with UNC was extended by Philips Medical Systems in 1992 [pdf], which enabled even more regular mutual visits between Utrecht and Chapel Hill (several per year). Together we wrote a highly cited paper (cited > 5270 times on Google Scholar) on Adaptive Histogram Equalization [Pizer1987]. Karel Zuiderveld generated the optimized code as a Graphics Gem.
A) Original image; (B) interactively windowed result; (C) unclipped AHE result; (D) clipped AHE result; CT abdomen. From [Pizer1987].
It was very inspiring to see the lab at UNC, experience the many innovations (head mounted displays, vibrating mirror 3D, ' PixelPlanes', shape statistics) and meet many extraordinary people, as prof. Henry Fuchs.
In those early days the Nuclear Medicine Department at AZU was also active with medical image processing for scintigrams (Kees de Graaf, Peter van Rijk) [Graaf1978].
We decided to join forces and established a working group "Image Processing in Medicine" with prof. Jan Koenderink and Ans van Doorn (Dept. Physics of Man), myself and Karel Zuiderveld (Dept. of Radiology), Stephen Pizer (UNC) and Max Viergever, who joined for a day/week from Delft University, and in 1988 was appointed as professor at Utrecht University, so we could continue as a university-based group.
In 1989 the Netherlands Ministries of Economic Affairs and Science & Education awarded a large grant to Utrecht University, to establish the national priority research program '3D Computer Vision'. This program was carried out by the Universities of Utrecht, Delft, Amsterdam and the University Hospital Utrecht. Meetings were held in 1990, 1991, 1992 ('3D Image Processing and Visualization') and 1993 ('Volume Image Procession', VIP' 93).
In 1993 we created the 'Image Sciences Institute' in Utrecht, which has grown ever since under the solid management of Max Viergever to the largest medical imaging institute in the Netherlands, see www.isi.uu.nl and [Viergever1999]. At that time, I was appointed associate processor and we both were the leading tenure staff members of the institute.
An overview of the many projects that we started:
In the late eighties Jan Koenderink wrote his seminal papers on differential geometry and multi-scale analysis ('scale-space theory'). I realized that, in order to build successful applications in medical image analysis, a fundamental physics approach was essential [see a nice overview in this paper: Haar1996]. Theoretical physics and mathematics were key inputs for such solid ground. In 1989 I managed to attract Luc Florack, a brilliant (cum laude) student from Theoretical Physics in Utrecht, as a PhD student. Later more theoretical physics came in, with Markus van Almsick, Stiliyan Kalitzin and Alfons Salden, deep mathmetics with Remco Duits, Joachim Weickerts, and many brilliant PhD students. This all helped the medical imaging applications, our core business, to be fundamentally well-designed, with always a keen eye on inspiration from our visual system: brain-inspired computing.
Luc was instrumental in founding our multi-scale differential geometric approach to (medical) computer vision. Luc is the best physicist I have ever met, and our life-long collaboration was instrumental for my career.
I was the supervisor of several PhD theses in Utrecht (with Max Viergever promotor):
Luc Florack received a cum laude PhD on 10-11-1993 on 'The syntactical structure of scalar images'. In 2001 Luc joined me to establish our new BMIA group at Eindhoven University. He became full professor in Mathematical Image Analysis at TU/e in 2007.
Bram van Ginneken received a cum laude PhD on 09-03-2001 on "X-thorax analysis for TBC screening". He later became full professor at Radboud University Nijmegen, where he founded and leads the Diagnostic Image Analysis Group. This group is now the largest hospital-based AI group in the Netherlands. He is CSO of the company Thirona.
Wiro Niessen received a cum laude PhD in 1997 on 'Multi-scale medical image analysis'. He has become full professor and head of the Biomedical Imaging Group Rotterdam and at Delft University of Technology. Active in population imaging, stroke, and founder of the compny Quantib.
Alfons Salden received a PhD on 12-11-1996 on 'Dynamic scale-space paradigms'. Alfons is a brilliant physicist, with a deep understanding of complex mathematics and an excellent physical intuititon.
I founded the international conference series 'Scale-Space' in 1989 in Utrecht. This bi-annual international conference was a few years later renamed into 'Scale-Space and Variational Methods' (SSVM).It has grown into an established computer vision event, with proceedings in Springer's Lecture Notes in Computer Science series and is still very active today (see the LNCS proceedings 1997-2021).
Here is the list of SSVM meetings so far: Utrecht, Netherlands, 1997 Corfu, Greece, 1999Vancouver, Canada, 2001Isle of Skye, UK, 2003Hofgeismar, Germany 200Ischia, Italy 2007Voss, Norway, 2009Ein Gedi, Israel, 2011Graz, Austria, 2013Bordeaux, France, 2015Kolding, Denmark, 2017Hofgeismar, Germany, 2019Cabourg, France, 2021 (virtual) Santa Margherita di Pula, Sardinia, Italy, 2023 Totnes, Devon, UK, 2025
At SSVM 2009 in Voss (Norway) I invited and introduced prof. Amiram Grinvald from the Weizmann Institute in Israel. He is the world's leading pioneer in voltage sensitive dyes, which enabled the revolution in optical imaging of the cortex, among which the discovery of the cortical 'pinwheels'. His work has featured on the cover of many journals.
Sample Text
Meeting so many colleagues again.
On the way home we made a beautiful trip through the Negev Desert.I rented a Bedouin tent for us, fun!
In 2011 I was co-chairing the Israeli edition of SSVM together with Freddy Bruckstein, and the twin brothers Alex and Michael Bronnstein. The location was Ein Gedi, a beautiful kibbutz at the Dead See,
In 2018 I co-chaired the ICIAR 2018 Conference in Povoa de Varzim, Portugal, with my colleague Aurelio Campilho .
Geometry-driven diffusion is an image processing technique to noise-clean images, while keeping good contours. It blurs the image locally, but adaptively not at the edges. It is the most used image processing technique today, working on any phone, TV, etc.
From 1992-1996 we were part of the European-American transatlantic collaboration 'Diffusion', sponsored by an joint ESPRIT-NSF grant. Six labs in Europe and 7 labs in the US were on board. The focus of the collaboration was non-linear anisotropic diffusion, and many key computer vision leaders participated. I edited the book of chapters that we all wrote as the fruits of our labour: 'Geometry-Driven Diffusion in Computer Vision [terHaarRomeny1994]. The wonderful preface was written by Jan Koenderink.
The collaborators and authors of the book chapters were (all professors now):Jitendra Malik, Pietro Perona, Tony Lindeberg, David Mumford, Ross Whitaker, Guido Gerig, Takahiro Shiota, Peter Olver, Guillermo Sapiro, Andrew Tannenbaum, Ben Kimia, Steven Zucker, Luis Alvarez, Jean-Michel Morel, Niklas Nordström, Luc van Gool, Marc Proesmans, Eric Pauwels, Antonio Leaci, Sergio Solimini, Thomas Richardson, Sanjoy Mitter, Luc Florack, Wiro Niessen, Alfons Salden, Max Viergever and myself.
The first two chapters in the Geometry-Driven Diffusion book I wrote with Tony Lindeberg, a pioneer in the scale-space world, now full professor at KTH in Stockholm. Here I visited him at his KTH office in January 2002.
We organized lively workshops at Berlin, Oxford UK, Stockholm and Palo Alto, see the pictures below.Virtually all of the members have become famous professors in computer vision. Prof. Jitendra Malik is with 222855 citations (Google Scholar) the world's most cited computer vision scientist today.
The main result of the research at Utrecht University was a multi-scale differential geometric model of early vision, which we used extensively for (bio-)medical image analysis applications. I wrote a monograph on 'Front-End Vision and Multi-Scale Image Analysis: Multi-Scale Computer Vision, Theory and Applications, written in Mathematica" (2003).
As the book is completely written in Mathematica, every discussed topic is accompanied by executable code, which greatly enhances understanding, and invites to 'play with the math'.
In 2003 I published the interactive book 'Front-End Vision and Multi-Scale Image Analysis, Multi-scale Computer Vision Theory and Applications, written in Mathematica' [terHaarRomeny2003]. It is one of the first computer vision books written as a çomputational essay, i.e. completely interactive. It has seen multiple reprints, and is used at universities worldwide.
I have been teaching the course 'Front-End Vision" annually from 1997-2019. This course was a 'base PhD course' for the research school 'Advanced School of Computing and Imaging' (ASCI), so virtually every PhD student in imaging in NL has been taking this course. This may have supported the relatively large growth of the medical image analysis field in the Netherlands.See for more details here.
FEV class of 1998. Upper right is Alex Frangi, famous for the 'vesselness paper' (cited 4448 times), now one of the top professors in Medical Image Analysis in the UK (see also here). Next to me Arjan Kuijper, professor at Fraunhofer and President of IAPR.More pictures in the TU/e section.
Our focus on solid mathematical physics and geometry attracted clever postdocs in our 'scale-space' group in Utrecht.
In 1996-1997 Joachim Weickert stayed as a postdoc with me and Luc Florack at Utrecht University, where he continued his work on anisotropic diffusion [Weickert1997}] He later became professor at Saarland University in Germany. In 2010, Joachim was awarded the Gottfried Wilhelm Leibniz Prize for his work in image processing (2.5 M€). It is the most important academic research prize in Germany. With over 3300 citations, his book 'Anisotropic Diffusion in Image Processing' [Weickert1998] (a re-write of his PhD thesis and work in Utrecht, Paris and Copenhagen) is one of the most cited monographs in the field of Mathematical Image Analysis.
From 1996-1998 Stiliyan Kalitzin, theoretical physicist, stayed as a postdoc in our group. He made great progress on our work on 'winding numbers', scale-space singular points and he layed the foundations for our later work on multi-scale multi-orientation analysis. He is currently the scientific leader of SEIN, the Foundation of Epilepsy Institutes in the Netherlands, in the city of Heemstede.
Fingerprint cleanup with coherence enhancing diffusion [Weickert 1999].