A combined post-filtering method to improve accuracy of variational optical flow estimation

Highlights：

• We propose a multi-scale and nonlinear structure tensor based 3D spatial-scale Harris edge detector for edges of flow fields.

• We propose a novel hybrid gradient bilateral and Gaussian filter approach through a spatial-scale gradient signal-to-noise ratio segmentation.

• We present a segmentation method to classify the structure tensor elements into continuity and non-discontinuity regions.

• A piecewise occlusions detection approach is used to detect occlusions of the flow field.

• We propose a combined post-filtering method with the weighted median filter, bilateral filter, and the fast median filter.