Set the motion estimating algorithm for the video stabilizer

stab.setMotionEstimator(motionEstType, ...)

stab.setMotionEstimator('KeypointBasedMotionEstimator', {estType, ...}, 'OptionName',optionValue, ...)
stab.setMotionEstimator('FromFileMotionReader', path, 'OptionName',optionValue, ...)
stab.setMotionEstimator('ToFileMotionWriter', path, {motionEstType, ...}, 'OptionName',optionValue, ...)

Input

• motionEstType Global 2D motion estimation methods which take frames as input. One of:
  • KeypointBasedMotionEstimator Describes a global 2D motion estimation method which uses keypoints detection and optical flow for matching.
  • FromFileMotionReader
  • ToFileMotionWriter
• path name of file for motion to read-from/write-to.
• estType Global motion estimation method, which estimates global motion between two 2D point clouds as a 3x3 2D transformation matrix. One of:
  • MotionEstimatorL1 Describes a global 2D motion estimation method which minimizes L1 error. Note: To be able to use this method you must build OpenCV with CLP library support.
  • MotionEstimatorRansacL2 Describes a robust RANSAC-based global 2D motion estimation method which minimizes L2 error.

Options

The following are options for the various algorithms:

KeypointBasedMotionEstimator, FromFileMotionReader, ToFileMotionWriter

• MotionModel Describes motion model between two point clouds. Default is based on the estimation method. One of:
  • Translation
  • TranslationAndScale
  • Rotation
  • Rigid
  • Similarity
  • Affine
  • Homography
  • Unknown

KeypointBasedMotionEstimator

• Detector feature detector, specified in the form: {detectorType, 'OptionName',optionValue, ...}. See cv.FeatureDetector.FeatureDetector for a list of supported feature detectors. Default is {'GFTTDetector'}.
• OpticalFlowEstimator sparse optical flow estimator specified as: {optflowType, 'OptionName',optionValue, ...}, where optflowType is one of:
  • SparsePyrLkOptFlowEstimator (default) wrapper around cv.calcOpticalFlowPyrLK.
  • SparsePyrLkOptFlowEstimatorGpu
• OutlierRejector outlier rejector specified as: {rejectorType, 'OptionName',optionValue, ...}, where rejectorType is one of:
  • NullOutlierRejector (default)
  • TranslationBasedLocalOutlierRejector

MotionEstimatorL1 and MotionEstimatorRansacL2

• MotionModel Describes motion model between two point clouds. One of:
  • Translation
  • TranslationAndScale
  • Rotation
  • Rigid
  • Similarity
  • Affine (default)
  • Homography
  • Unknown

MotionEstimatorRansacL2

• MinInlierRatio default 0.1

• RansacParams Describes RANSAC method parameters. A struct with the following fields:

  • Size Subset size.
  • Thresh Maximum re-projection error value to classify as inlier.
  • Eps Maximum ratio of incorrect correspondences.
  • Prob Required success probability.

  If a string is passed, it uses the default RANSAC parameters for the given motion model. Here are the defaults corresponding to each motion model:

  • Translation struct('Size',1, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • TranslationAndScale struct('Size',2, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Rotation struct('Size',1, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Rigid struct('Size',2, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Similarity struct('Size',2, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Affine struct('Size',3, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Homography struct('Size',4, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)

  By default is it set to 'Affine'.

SparsePyrLkOptFlowEstimator

• WinSize Size of the search window at each pyramid level. default [21,21]
• MaxLevel 0-based maximal pyramid level number. default 3

TranslationBasedLocalOutlierRejector

• CellSize default [50,50]

• RansacParams Describes RANSAC method parameters. A struct with the following fields:

  • Size Subset size.
  • Thresh Maximum re-projection error value to classify as inlier.
  • Eps Maximum ratio of incorrect correspondences.
  • Prob Required success probability.

  If a string is passed, it uses the default RANSAC parameters for the given motion model. Here are the defaults corresponding to each motion model:

  • Translation struct('Size',1, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • TranslationAndScale struct('Size',2, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Rotation struct('Size',1, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Rigid struct('Size',2, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Similarity struct('Size',2, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Affine struct('Size',3, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)
  • Homography struct('Size',4, 'Thresh',0.5, 'Eps',0.5, 'Prob',0.99)

  By default is it set to 'Translation'.

The class uses KeypointBasedMotionEstimator by default with MotionEstimatorRansacL2.