next up previous contents
Next: Default stereo matching Up: Stereo Correlation Matching Previous: Correlate.   Contents

disp drad.

Having obtained a disparity image from typically Edges or Corners the data can be viewed using Display 3D. The 3D projection of the data may contain points which are visibly incorrect. A filter for the incorrect data which uses the disparity gradient limit can be applied to the disparity image. The parameters of Disp Grad are accessible via the Disp Grad Params dialogue box.

disp grad params

The following parameters, used by Disp Grad, can be accessed by this dialogue box:

Limit. The smaller this value is, then the larger the amount of rejected data. Psycho-physics research suggests that a value of 1.0 is used in the human vision system as a limit for stereo fusion, and a value of greater than 2.0 along an epipolar represents order reversal.

Must Pass %. The disparity gradient test is binary with respect to the imposed limit, the percentage of the surrounding points with which the data point in question must pass is defined with this value. In normal scenes a point will not be expected to pass the disparity gradient test with 100 Nearest N. This number defines the N nearest points on which the disparity gradient test will be applied. The search path is a rectangular spiral outwards from the data point in question.

When setting up these parameters a tradeoff must be found between removing all of the incorrect points, which fail the disparity gradient test, and removing to many correct points which also violate the disparity gradient test. Bearing in mind that experimental error is increased both by including incorrect data and by discarding correct and potentially useful data points.


next up previous contents
Next: Default stereo matching Up: Stereo Correlation Matching Previous: Correlate.   Contents
root 2017-09-24