Disparity estimation using block matching and dynamic programming

<title>Disparity estimation hardware for real-time stereoscopic applications</title>

“…So, every point P. that is visible from both cameras, with coordinates (x,y,z) can be represented from the left view with coordinates (x1 , y1) and the right Bt.ttiint LOft Ctttt Right Ct.mc,t view with coordinates (Xr , Yr ). Representing as f the focal length of the two cameras and using the similar triangles theorem we can find that: xl x+d/2 11= z (1) xx-d/2 (2) f z yl Yr Y -=-=- (3) ffz Examining equations (1), (2) and (3), we observe that the point with coordinates (x,Y,z) is represented with different x-coordinates in the two image views. The difference x1 -xr is commonly called the disparitY of the two views.…”

“…Many algorithms have been proposed [2][3][4][5][6] for the disparity estimation problem, but most of them can be used for software off-line implementations. In addition, the proposed disparity estimation algorithms find difficulties during the disparity calculation of invisible areas from either camera.…”

Stereo image coding with disparity compensation using dynamic programming

“…Recently, combination of block matching and dynamic programming is proposed to estimate disparity and calculate intermediate view [2].…”

“…where -dmaX<6dmax , a is smoothing parameter [ 2 ] . The term dc(t,m) represents the cost of assigning the neighboring blocks (ij) and (i,j+l) of the left image to the blocks displaced by m pels and t pels in the right image, respectively.…”

“…This can be improved by feeding all possible correspondences with their matching costs to a dynamic programming algorithm and find a path which minimizes the total matching cost along the path [2]. But if occlusion occurs, due to finite viewing area or depth discontinuity, disparity estimation and compensation will fail.…”