Research Article

Epipolar Plane Image Rectification and Flat Surface Detection in Light Field

Algorithm 1

The proposed strategy for flat surface detection.
() Input: Light field LF, slope range , the
  number of training samples , the number of testing
  samples
() Output: The energy ratio of the feature points which
  have the same slope value with in all feature points
  which come from all rectified EPIs.
() / Plane fitting stage /
() Choose EPIs from LF randomly.
() Calculate the best parameters of each EPI by
  using the algorithm mentioned in Section 3.2.
() Select the which has the most frequency in all as
  the common . And fit the linear function of with
  by using the whose equals to .
() / Feature extracting stage /
() Choose EPIs from LF randomly.
() Calculate the parameters of each EPI by using
  the common and the linear function of with .
() Rectify each EPI with its by using the algorithm
  mentioned in Section 3.1.
() Extract the Lisad-1 feature points of all rectified EPIs.
() Count the ratio of the feature points which have
  same slope value with for each EPI, and contruct a
  the feature vector by combining these ratios in a
  descending order.
() Employ a classifier to distinguish flat surface from
  natural scenes.