This paper combines the reactive collision avoidance methods with image-based visual servoing control for mobile robot navigation in an indoor environment. The proposed strategy allows the mobile robot to reach a desired position, described by a natural visual target, among unknown obstacles. While the robot avoids the obstacles, the camera could lose its target, which makes visual servoing fail. We propose in this paper a strategy to deal with the loss of visual features by taking advantage of the odometric data sensing. Obstacles are detected by the laser range finder and their boundaries are modeled using B-spline curves. We validate our strategy in a real experiment for an indoor mobile robot navigation in presence of obstacles.