Unsupervised Markovian Segmentation on Graphics Hardware

Hierarchical Markov Random Fields Applied to Model Soft Tissue Deformations on Graphics Hardware

“…Monte-Carlo optimization). Secondly, the nature of the Markovian approach resulted in a straightforward implementation in the GPU, which has been remarked by others [9,10]. The experiments showed that by using HRMF at the GPU level, stable results can be attainable in a very reasonably computation time, simplifying the soft-tissuedeformation simulation pipeline.…”

“…In [9,10] the general way of MRF on graphics hardware is presented. For the implementation a new extension for OpenGL, EXT_framebuffer_object [6,7], was used.…”

“…The hierarchical approach is also robust with respect to local minima [14]. Finally, MRF and HMRF are well suited for parallel implementation [9,10], which has brought more attention with the advent of GPU-based application programming interfaces (API).…”

Detecting Half-Occlusion with a Fast Region-Based Fusion Procedure

“…As for the implementation, since every pixel of the scene are independently processed, we have implemented our method on a parallel architecture, namely a Graphics Processor Unit (GPU) [17]. A GPU is a processor embedded on most graphics card nowadays available on the market which, among other things, can load, compile and execute programs implemented with a C-like language.…”

“…To our opinion, this property appears as a major advantage. Second, this segmentation method can be parallelized and implemented on a parallel architecture such as a graphics processor unit (GPU) [17]. In this way, the segmentation map r c can be computed in interactive time.…”