/Optical-Flow

Implementation of Lucas-Kanade optical flow algorithm.

Primary LanguageMatlab

Implementation of Optical Flow Algorithm

The implementation has 4 parts:

  1. Naive dense optical flow.(opticalFlow.m) Windowsize and threshold for smallest eigen value are free parameter

  2. Corner-based sparse optical flow

  • Corner detection is based on Gaussian deviation (CornerDetect.m, gaussian.m, d_gaussian.m)
  1. Iterative Coarse to Fine Optical Flow (details can be found in report.pdf)
  • Multi-resolution pyramid (pyramidFlow.m)
  • apply Lucas-Kanada optical flow iteratively to estimate potential motion velocity on each level (iterOpticalflow.m)
  1. Experiments
  • Motion based Background Subtraction (bg.m) Remove the dynamic portions of an image from a static background
  • Motion based Image Segmentation