Motion Estimating Optical Flow for Action Recognition : (FARNEBACK, HORN SCHUNCK, LUCAS KANADE AND LUCAS-KANADE DERIVATIVE OF GAUSSIAN)

Abstract

Motion estimating is one of the methods which determines the movement from one frame to another in the videos. For an application of action recognition, choosing the optical flow can be an essential feature for recognizing actions. The optical flow consists of the information of the moving subject and objects in the video frames. This paper analyzes four motion estimating optical flow methods (Farneback, Horn Schunck, Lucas Kanade, and Lucas-Kanade Derivative of Gaussian explored based on visualization and PSNR. The NTURGB+D dataset uses for the analysis of experimental results.