Research
Image processing: Partial Differential Equations and Variational Approaches
Partial differential equations and variational approaches were introduced into image processing in the 90s. One of the main interests of this formalism is that the theory is well established. Intensive research has been carried out in this area, and I contributed in the following domains:
Image restoration and enhancement via partial differential equations: icip-97, siam-09
Image super-resolution in fMRI: ijist-04, miccai-03
Image inpainting: rr-06
Interferometric phase image restoration and unwrapping: icpr-02, icip-02
Optical flow estimation using variational approaches: siam-99, siam-99b
Optical flow estimation using the structure tensor: bmvc-04
Transparent motion estimation: jmiv-11
Video segmentation: jmiv-99
(see also our book on this topic)
Retina Modeling and Applications
The retina is far to be a simple linear analogue to digital converter: Not only the retina allows to capture a variety of rich behaviors but also the information encoded in the spike trains cannot simply be interpreted by just counting the spikes. My interest in this area are to (i) design bio-inspired retina models that can reproduce behaviors observed in real cells recordings, (ii) interpret the spiking output with the perspective of proposing novel approaches for computer vision problems (e.g., coding/decoding algorithms, categorization):
Model of the retina: jcn-09, soft
Bio-inspired coding/decoding schemes for images and videos:
icassp-10,
acivs-11,
tnnls-12,
rr-12
Bio-Inspired Motion Estimation and Applications
In the field of computer vision, motion estimation is still an active and challenging field of research. Meanwhile, research in neuroscience and psychophysics investigates how the brain is "making" our motion percept. My interest is to bridge the gap between the two domains, by proposing novel bio-inspired models of motion estimation, based on the cortical architecture, that reproduce major perceptual phenomena while being efficient to solve computer vision problems.
Bio-inspired dynamical models of motion integration: vr-10, eurasip-11
Neural fields of motion integration and bifurcation analysis: mip-11, rr-11
Bio-inspired benchmark for motion estimation:
bimv-10,
website
Action recognition:
cviu-12,
ijcv-09, eccv-08
Other Computer Vision Topics
Some other topics I worked on are listed below:
Tensor voting: cvpr-00
Reading-aid systems for low vision patients: cvavi-08
Interest point detectors: cvpr-09
Saliency estimation in videos: icip-09
Bilateral filtering: tcgv-10
Bag-of-features and image classification: kdir-11
