Voting Algorithm

• one possible segmentation algorithm
• Algorithm:
  When a feature vector is computed, the associated label that is derived is usually applied to the pixel at the center of the window chosen for the calculation. The voting method builds on this by noting that each pixel is used in many different windows. Therefore, each window, upon deciding on a label, can ``cast a vote'' at each pixel in its window. After all computations have been carried out, each pixel will contain a certain number of votes for each possible label, and the label with the most votes becomes the label for the pixel. This requires more computations than are used in split-and-merge, but the resulting regions have boundaries that appear to be smooth rather than blocky.