Determination of objects contours using physical principles
Patrícia C. T. Gonçalves, João Manuel R. S. Tavares, Renato Natal Jorge · 2008
Segmentation, the identification of an object represented in a static image or along image sequences, is one of the most common and complex tasks in the domain of Computational Vision. Usually, whenever we intend to extract higher level information from images, we need to start by segmenting them. The main goal of this work is to segment an object represented in an image by extracting its contour after defining an initial contour for it; this coarse contour will evolve along an iterative process until it reaches the frontier of the desired object, figure 1. For that purpose, a deformable model is used, whose behaviour is driven by physical principles. Initial contour Final contour Figure 1. The main purpose of this work: starting from a coarse shape evolve to one that defines well the contour of the desired object. More precisely, in our methodology, the initial contour is transformed into an elastic physical model – assigning mass, rigidity and damping to each of its points – that is gradually deformed in order to