Skeletonizing Digital Images with Cellular Automata - Cellular Automata in Image Processing and Geometry - page 58

Fig. 3.11 A comparison of our implementation vs. the sequential one by taking 90 steps in
the Guo and Hall algorithm

3.6 Conclusions

Computer vision is a hard task and a challenge in the next years. Classical sequential
algorithms need to be revisited and adapted to the novel technologies, but the new
developments also need the support of deep theoretical foundations. On the one
hand, it is necessary to develop new algorithms in the framework of CA or other
computational paradigms which can put a new light on classical problems. A deep
study of the properties of such algorithms can improve their practical efficiency and
accelerate their use in digital image industry.
On the other hand, the inherent features of CA for dealing with Image Analysis
have found the limits of sequential computers. The theoretical parallel framework of
CA could not be efficiently implemented in one-processor computers. In this way,
the theoretical study should be also oriented to the most recent advances in hardware
architecture. In this chapter, we have considered the implementation on GPUs. It is
a good alternative for future implementations for CA algorithms, but many other
possibilities should be explored. A strong link between the research on theoretical
models and the development of new hardware architectures is the key for realistic
answers to the challenges of the future in digital image areas.

Acknowledgements. MAGN acknowledges the support of the project TIN2012-37434 of the
Ministerio de Economía y Competitividad of Spain.

References

1. Acerbi, L., Dennunzio, A., Formenti, E.: Conservation of some dynamical properties for
operations on cellular automata. Theoretical Computer Science 410(38-40), 3685-3693
(2009)