Computer vision: Optical Flow

All Optical Flow implementations can process only files in pgm format. In order to visualize the optical flow the next color scheme is used:

It contains:

• Lucas_Kanade - Lucas-Kanade optical flow.The optical flow based on gray value constancy assumption.

  • To install use in terminal:
    • cd ../cv_opticFlow/Lucas_Kanade
    • make
  • To run: ./Lucas_Kanade name_of_folder_with_picture_sequence

• Horn_Schunck - Horn_Schunck optical flow based on variational methods.

  • To install use in terminal:
    • cd ../cv_opticFlow/Horn_Schunck
    • make
  • To run: ./Horn_Schunck name_of_folder_with_picture_sequence

• Horn_Schunck_TV - Horn_Schunck optical flow based on variational methods with total variation (TV) smoothness term (robust penalizer for sharp edges).

  • To install use in terminal:
    • cd ../cv_opticFlow/Horn_Schunck_TV
    • make
  • To run: ./Horn_Schunck_TV name_of_folder_with_picture_sequence

• Horn_Schunck_TVbothTerms - Horn_Schunck optical flow based on variational methods with total variation (TV) both terms.

  • To install use in terminal:
    • cd ../cv_opticFlow/Horn_Schunck_TVbothTerms
    • make
  • To run: ./Horn_Schunck_TVboth name_of_folder_with_picture_sequence

• Horn_Schunck_TVbothTerms_Grad - Horn_Schunck optical flow based on variational methods with total variation (TV) both terms plus gradient constancy assumption which makes optical flow robust to illumination changes.

  • To install use in terminal:
    • cd ../cv_opticFlow/Horn_Schunck_TVbothTerms_Grad
    • make
  • To run: ./Horn_Schunck_TVbothTerms_Grad name_of_folder_with_picture_sequence

• Brox_2004 - Thomas Brox large displacements optical flow based on variational methods without gradient constancy assumption.

  • To install use in terminal:
    • cd ../cv_opticFlow/Brox_2004
    • make
  • To run: ./Brox_2004 name_of_folder_with_picture_sequence

• denoiz - filtering algorithm which is equivalent to Gauss blur filtering but it based on the energy minimization approach. It could be helpfull for improving quality of the optical flow for noisy images.

  * To install use in terminal: 
    * cd ../cv_opticFlow/denoiz
    * make
  * To run: ./denoiz name_of_file