Algorithm 1 Generate points of interest and possible patches
Require:	Points of interest
	(a) Obtain user input bounding box during the first frame
	(b) Convert previous and current frames to grey scale for each cycle
	(c) Perform Smoothing to both frames with Gaussian convolution
	(d) Find Laplacian for each frame
	(e) Obtain the matrix $\left[\begin{array}{cc}{F}_{I_{xx}}^{i,x,y}& F_{I_{xy}}^{i,x,y}\\ F_{I_{yx}}^{t,x,y}& F_{I_{yy}}^{t,x,y}\end{array}\right]$ for each pixel
	(f) If min eigenvalue>
	accept the point as possible POI
	Else
	reject the point
	End If
Require:	tentative patches
	(a) Each channel of RGB is treated separately
	(b) Built vector descriptors, ${\mathbf{\text{V}}}_i^{t,x,y}$ for both frames
	(c) For i, …, total POI Do
	$L_1={\sum }_{\forall i}|{\mathbf{\text{V}}}_i^{t,x,y}-{\mathbf{\text{V}}}_i^{t-1,x,y}|$
	End For
	(d) Obtain L2 to decide either to accept the POI or not
	If L1 < 1
	$L_2^{t,x,y}=1$
	Else
	$L_2^{t,x,y}=0$
	End If
	(e) Construct a patch at each point where L2 = 1
	The original size of each patch is similar with the final size of the previous frame
	(d) Smooth out overlapping patches by combining redundant patches
	IF Oa > ( 2 × βw × βh)
	combine the patches by finding the mean of the patches
	Else
	remain as it is
	End If