correlation based disparity map on opencv result is not satisfying enough -
i have used normalized cross correlation equation find disparity between 2 rectified images of tsukaba.

i searched in horizontal direction in range x=-15:1:15 , in vertical direction in range y=-1:1:1 using rectangular windows sized 9x9.
then (x,y) maximized correlation, disparity (i,j) calculated : disparity(i,j)=sqrt( x^2 +y^2 ).
this disparity image get:

which not satisfying ground truth disparity map:

is there post processing have perform enhance result?
thanks in advance.
Comments
Post a Comment