Traffic safety is occupying a prominent place in the agenda of cars manufacturers and suppliers. Their aim is to develop reliable systems which recognize and anticipate possible hazard situations. One of such systems is called Advanced Driver Assistance Systems (ADAS). One of the principal components of image based ADAS is the motion estimation of traffic participants (visual object tracking). Object tracking estimates the trajectory of an object by labeling its position in every frame as it moves around a scene. Object tracking is commonly confronted with a variety of unfavorable visual conditions typical from outdoor scenarios such as rain, darkness reflections, illumination variance, occlusions, etc., where classical tracking algorithms as Lucas-Kanade show some limitations. Therefore, the relevance of visual tracking for the correct operation of ADAS inevitably entails the use of a robust and reliable algorithm, which shows outstanding performance for outdoor scenarios under typical constraints for automotive applications such as low complexity, real time operation, low memory consumption, reliability and robustness. The main goal of this thesis is to investigate and compare the state of the art of visual tracking to find a robust and reliable algorithm, which overcomes challenging outdoor scenarios and exhibits comparable or better performance than algorithms based on classical approaches as Lucas-Kanade, under the above-mentioned typical constraints for automotive applications. As result of the research and carried out experiments with typical automotive scenarios, the Spatially Regularized Correlation Filter presented by Danelljan et al. in [DHSF15b] was selected and extended with some proposed improvements. Conducted experiments show the applicability of the algorithm and outstanding performance as well for challenging outdoor scenarios as for controlled scenarios both under automotive constraints. Significant improvements based on the Kalman filter were proposed to enrich the accuracy and robustness of the filter. The experiments realized with the proposals demonstrate an improvement on the results of the filter for all the tested sequences. It can be concluded that the original goal of the thesis was fulfilled, and the selected algorithm with the implemented proposals constitutes an optimal solution to the presented challenges in object tracking in realistic ADAS scenarios.