A segmentation method based on motion from image sequence and depth

The author proposes a segmentation method based on the gradient method for multiple moving objects which may include an object for which a unique interpretation of the motion is difficult. By the gradient method, the 3-D motion parameters of a rigid object can be estimated without determining correspondence as a pseudoinverse solution for a system of linear equations if the 3-D structure of the object is already given. Based on the residual square-error in the motion estimation, the image is segmented if it contains regions with different motions. If the motions are recognized as the same, the regions are merged. Thus, the object is extracted by iterating the segmentation and merge on the image