In this work, we present a fast approach to estimate the motion parameter coefficients,which results in a significant reduction of the computational time of the 3D motionsegmentation approach as well as a decrease in the mean error of the estimated parameterseven with highly noisy MVF. Furthermore, a saliency-based approach forestimating and segmenting 3D motions of multiple moving objects represented by2D motion vector fields (MVF) was developed. A classification module has been implementedto define the global motion of the mounted camera in order to overcometypical problems in autonomous mobile robotic vision such as noise, occlusions, andinhibition of the ego-motion defects of a moving camera head. Moreover, we proposea fast depth-integrated 3D motion parameter estimation approach which takes intoconsideration the perspective transformation and the depth information to accurately estimate biologically motivated classifier cells in the 3D space using the geometricalinformation of the stereo camera head. The results show a successful detection andestimation of predefined 3D motion patterns such as movements toward the robotwhich is a vital milestone towards a successful prediction of possible collisions.