python - 如何让 cv2.calcOpticalFlowFarneback() 更有信心

Question

我做了一个小测试cv2.calcOpticalFlowFarneback()来检查它的结果有多稳定。在我的测试中，我用相机拍了一张快照，剪切了两个重叠的图像，(0,10)它们之间有距离。我对每个像素都是预期(0,10)的，但结果却大不相同。

import cv2

cap = cv2.VideoCapture(0)
ret, img = cap.read()
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
cv2.imshow('gray', gray)

left = gray[:-5, :]
right = gray[5:, :]

flow = cv2.calcOpticalFlowFarneback(left, right, 0.5, 3, 15, 3, 5, 1.2, 0)
flowx = flow[:,:,0]
flowy = flow[:,:,1]

cv2.imshow('flowx', flowx)
cv2.imshow('flowy', flowy)

print 'x.mean:', flowx.mean()
print 'x.max:', flowx.max()
print 'x.min:', flowx.min()

print 'y.mean:', flowy.mean()
print 'y.max:', flowy.max()
print 'y.min:', flowy.min()

cv2.waitKey()

结果：

x.mean: -0.00440598979749
x.max: 1.53141
x.min: -1.95973
y.mean: -2.53464987664
y.max: 0.844686
y.min: -5.43879

我怎样才能使它更接近我的预期(0,10)结果？