Professor Moulin's research involves the development of novel methods for modeling and processing signals, images, and video, with a focus on problems of compression, restoration, and, more recently, information hiding and authentication. Applications of interest include videoconferencing, digital TV, multimedia services, and computed imaging. A problem that is fundamental and common to those applications is the need to find sparse signal representations and processing techniques that optimize appropriate performance measures. The work aims to broaden the scope of information-theoretic methods and results on optimal signal processing that currently apply to somewhat narrow classes of idealized signals.
Much of Prof. Moulin's work involves wavelets and related multiresolution techniques, including designs adapted to signal statistics. In image compression, he has developed optimal quantization techniques for nonorthogonal filter banks and applied these techniques to image and video coding. Together with his coworkers, he has also developed efficient techniques for modeling, estimation, and coding of dense video motion fields. The techniques are a promising alternative to the standard block-based motion estimation techniques. His work in statistical signal and image processing covers spectral density estimation, radar imaging, complexity-regularized image restoration, and Bayesian image restoration.
Since late 1997 he has been developing an information theory of watermarking and data hiding, together with J. A. O'Sullivan (Washington U.).