Contact FutureLearn for Support Skip main navigation
We use cookies to give you a better experience, if that’s ok you can close this message and carry on browsing. For more info read our cookies policy.
We use cookies to give you a better experience. Carry on browsing if you're happy with this, or read our cookies policy for more information.

Skip to 0 minutes and 6 secondsWell Thomas, in this course, we've seen a lot of the work that's being done in your research group. What is your interest in shape models why do you think they are important? So the first answer is clearly related to this course. As we have seen, shape modelling is extremely important for medical image analysis. But there's also a second part. We work on 2D images, mainly on faces. And there we also use 3D shape models to analyse photographs. And what exactly do you do with the 2D images? So I said we work mainly on faces and, for instance, we can reconstruct the 3D shape of a face from a single photograph.

Skip to 0 minutes and 48 secondsAnd then through this reconstruction we can apply transformations that we learned earlier, such as ageing.

Skip to 0 minutes and 57 secondsAnd then we can re-project this transformation and age a photograph.

Skip to 1 minute and 8 secondsWow that's amazing.

Skip to 1 minute and 11 secondsAnd why was this work and 2D images not included in the course? Well 2D image analysis is more difficult than it might appear. So in principle, it's very similar to fitting a shape model to the intensities of an MRI images. But the fitting process for images is a bit more complicated mainly since the intensities of the images are not only given by the surface of the face, it's also the colour of the illumination, especially the direction of the illumination. And why is this illumination, colour and pose such a problem and how can you solve it? If you take, for example, face recognition and the varying pose in illumination, still a very difficult problem for computers.

Skip to 1 minute and 59 secondsThe photographs of a single face can be very different. So our approach for this compensating illumination and pose is that we reconstruct the face in an image with our 3D model, and then instead of comparing the 2D images, we compare our reconstructions. And the reconstructions are way more independent of pose and illumination. OK. We've seen face recognition and ageing which was pretty amazing. Are there any other examples of your work on 2D images and faces you could show us? Of course you could also do some fun stuff such as making the ladies and gentlemen in all the paintings come alive.

Shape modelling for the analysis of 2D images

And here is yet another application of shape modelling: the reconstruction of 3D face models from 2D photographs.

In this interview, Professor Vetter explains why this is an important task and how shape models can help to solve it. He then sheds light on the difficulty of modelling illumination – a main challenge in the analysis of photographs and the reason why the topic of 3D reconstruction of 2D images was not included in the course. Further, you will see Antonia Bertschinger, the interviewer, suddenly age in an alarming fashion, and you will watch a painting come alive.

This video is from the free online course:

Statistical Shape Modelling: Computing the Human Anatomy

University of Basel