Source: ACM SIGGRAPH Citation
Eduardo’s PhD work introduced three novel and efficient ways of looking at and understanding High-Dimensional Filters (HDF). Such operators arc a fundamental building block for many image and video processing applications, including detail manipulation, denoising, tone mapping, photon-mapping filtering, upscaling, spatio-temporal filtering, rocoloring, and stylization, just to name a few. Prior to his work, high-dimensional filtering techniques were able to produce great results in many practical situations. but were not computationally efficient in all scenarios. Eduardo’s dissertation introduced three optimal and parallelizable solutions for different aspects of HDF, all of them possessing linear-time complexity in both the number of samples being filtered and in the dimensionality of the space in which the filters operate.
The first contribution, known as Domain Transform, offers an efficient solution for performing high-dimensional (an)isotropic filtering using a geodesic metric. It was introduced at SIGGRAPH 2011 and is among the fastest geodesic HDF techniques available both for CPU and GPU.
The second contribution, called Adaptive Manifolds, introduced an efficient solution for (an)isotropic HOF using a Euclidean metric. This work also provided the first demonstration or a singlepass hybrid Euclidean-geodesic filter. The Adaptive Manifolds technique was introduced at SIGGRAPH 2012. Like the Domain Transform, it is among the fastest Euclidean HDF techniques available for both CPU and GPU.
The third contribution is a discrete-time mathematical formulation for applying arbitrary recursive filters to non-uniformly sampled signals. This formulation enabled, for the first time, geodesic edgeaware evaluation of arbitrary recursive infinite impulse response filters (not only low-pass), which allows practically unlimited control over the shape of the filtering kernel. This technique was introduced at EUROGRAPHICS 2015, where it received a Best Paper Award Honorable Mention.
Eduardo Simoes Lopes Gastal received his BSc (2010) and PhD (2015) degrees in Computer Science from the Federal University of Rio Grande do Sul (UFRGS), in Brazil, under the supervision of Manuel M. Oliveira. He is currently a Postdoctoral Fellow at the Institute of Informatics at UFRGS.