Ogura Takeshi

Evaluation of advanced pixel-level snakes on cellular hardware platform

In this paper we implemented pixel level snakes on cellular hardware platform. Cellular hardware platform can process cellular structural computation paradigm, which includes Cellular Automaton (CA), Cellular Neural Network (CNN), and so forth, in parallel. One of implementations of cellular hardware platform is Cellular AutoMata on Content Addressable Memory (CAM2), which has Content Addressable Memory (CAM) based architecture, and we use it for the implementation of pixel level snakes. In our implementation, CNN, which played important roles in the original literature, were substituted by CA for reason of complexity. Our proposed implementation method of pixel level snakes was simulated on an instruction set level hardware simulator, and the validity of our method was confirmed. The estimated time of contour extraction by contract process is about 493ms for 128 × 128 pixel image.