A Cellular Active Contours Algorithm Based on Region Evolution

Abstract

This paper introduces a new algorithm for implementing cellular active contour technique based on pixel-level snakes (PLS). The main motivation is the optimization of the computational performance, especially when PLS are implemented on pixel-parallel single instruction multiple data (SIMD) processor arrays. The algorithm is based on the evolution of an active region. This allows the implementation of the entire algorithm using every simple local rules. Additionally, nested contours and propagation into narrow cavities are supported. The algorithm and experimental results from real-time implementation on vision chips are presented