TGSF/TLoG filter with optical flow technique for large motion detection

In this paper, detection and segmentation of large motion in moving image sequences is presented. For detecting motion, the intensity of each pixel is convolved with the second derivative of the Temporal Gaussian Smoothing Function (TGSF) or the Temporal Laplacian of Gaussian (TLoG) filter. The zero-crossing in a single frame of the resulting function indicates the positions of moving edges. An intensity change over time due to a small illumination effect does not produce a zero crossing. Therefore, such changes are not interpreted as human motion by this method. The optical flow velocity is computed by using the spatial and temporal derivatives of this function, and it is normal to the zero crossing contours. Pixels belonging to the normal velocities are projected back to the original color image sequences to achieve a segmented color image. Experiments show that a moving object is detected correctly, and good segmentation results are achieved.