Computer Vision News - September 2022

32 Best Paper WBIR 2022 Weighted Metamorphosis is a modification of the Metamorphosis framework introduced by Alain Trouvé and Laurent Younes . It is an extension of Large Deformation Diffeomorphic Metric Mapping (LDDMM) , allowing local intensity changes. Therefore, it enables the registration of two images with diffeomorphic deformations that do not have the same topology. LDDMM is a widely studied algorithm and framework, but Anton tells us it is not yet very well understood by many people working within the field of image registration, and Metamorphosis is even more complicated. Its origins date back almost two decades, but it has not seen anynotableapplicationor implementation since then. “ We figured out that Metamorphosis didn’t work as well as we expected, ” Anton says. “ It is a non-machine learning type algorithm – everything is modeled. In the paper, we have this first equation where we have the parameter balancing deformation and intensity changes to WEIGHTED METAMORPHOSIS FOR REGISTRATION OF IMAGES WITH DIFFERENT TOPOLOGY Anton François is a PhD student at Paris Cité University, under the supervision of Joan Alexis Glaunès and Pietro Gori from Télécom Paris. He speaks to us fresh from winning the Best Paper award at WBIR 2022 for his work on Weighted Metamorphosis. WBIR 2022 Best Paper