One of the major problems encountered during an automatic contour detection is the lack of structure continuity perception. Active contours are currently used to represent this continuity, but they point out some problems for contour detection in textured, noisy and low contrast images. In this paper, we present a new approach to perceive continuity based on a 2D reconstruction of closed contours using a multi-agent system. Each agent is provided with sensors on the image, which allow it to follow local extrema. The purpose is to detect alternative light and dark concentric structures in an image. In the first version of the method, the closed contour was validated when the agent had found again its initial position after having described a circular path. Actually, this condition was not easily satisfied in bad contrasted images. We have added high-level information about the shape of the contour to improve the detection. An application to fish otolith growth ring detection is presented in this paper.