“Substituting crossed textons for super resolution of video” by Kamimura, Tsumura, Motomura and Kanamori

  • ©Kenji Kamimura, Norimichi Tsumura, Hideto Motomura, and Katsuhiro Kanamori




    Substituting crossed textons for super resolution of video



    It is difficult to achieve real-time processing using the conventional example-based super-resolution method, since it requires high computational cost [Freeman et al. 2002]. In contrast, real-time processing can be achieved using the texton substitution method we proposed previously. Since the textons are defined at each pixel, their substitution can be accelerated using parallel processing hardware (e.g. GPUs). However, this method tends to result in noisy images, since the relationships between high and low resolution textons are not always unique. To resolve this problem, in this paper, we consider the spatial connections of textons for uniquely substituting textons into a texton. The results demonstrate that the output images have improved quality without any increase in the calculation cost.  


    1. Freeman, W. T., Jones, T. R., Pasztor, E. C., Example-Based Super-Resolution, IEEE CG&A 2002, 22, 2, 56–65.

ACM Digital Library Publication:

Overview Page: