Wavelet Shrinkage in Optical Flow Estimation
shrinkage; denoising; estimation
Optical Flow Computation is the estimation of 2D image motion, a projection of 3D motion onto its image plan. It uses a time-ordered image sequence to calculate the instantaneous velocities at image points. The image sequences are always assumed to be noiseless in the computation of optical flow. However, it is not a realistic assumption in real life. In this paper, we investigated the motion of rigid object(s) using Horn and Schunk's differential method to estimate the optical flow in image sequences contaminated with additive white Gaussian noise. We carried out wavelet shrinkage (denoising) using different thresholding methods. It is shown that wavelet shrinkage can be a very effective tool in restoring the "true" optical flow if the nosie model of the motion vector is known.
MetadataShow full item record
- ECE Publications