Hassan Mostafavi and Fred W. Smith. Image correlation with geometric distortion. Part II: Effect on local accuracy. IEEE Transactions on Aerospace and Electronic Systems, AES-14(3):494--500, May 1978.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....and magnification) Correlation is translation invariant but requires that template and image have the same scale and orientation. If this is not the case, the correlation value deteriorates, even if the unscaled and unrotated template is identical with the image. In [Mostafavi and Smith, 1978a; Mostafavi and Smith, 1978b] it is shown that in the case of an affine transformation between image and reference, there are template sizes and template shapes that minimize the probability of a mismatch as well as the registration error. The more severe the distortion is, the smaller the optimal template size becomes. Many ....

Hassan Mostafavi and Fred W. Smith. Image correlation with geometric distortion. Part II: Effect on local accuracy. IEEE Transactions on Aerospace and Electronic Systems, AES-14(3):494--500, May 1978.

....noise, blurring, and magnification) Correlation is translation invariant but requires that template and image have the same scale and orientation. If this is not the case, the correlation value deteriorates, even if the unscaled and unrotated template is identical with the image. In [Mostafavi and Smith, 1978a; Mostafavi and Smith, 1978b] it is shown that in the case of an affine transformation between image and reference, there are template sizes and template shapes that minimize the probability of a mismatch as well as the registration error. The more severe the distortion is, the smaller the ....

....and magnification) Correlation is translation invariant but requires that template and image have the same scale and orientation. If this is not the case, the correlation value deteriorates, even if the unscaled and unrotated template is identical with the image. In [Mostafavi and Smith, 1978a; Mostafavi and Smith, 1978b] it is shown that in the case of an affine transformation between image and reference, there are template sizes and template shapes that minimize the probability of a mismatch as well as the registration error. The more severe the distortion is, the smaller the optimal template size becomes. ....

Hassan Mostafavi and Fred W. Smith. Image correlation with geometric distortion. Part I: Acquisition performance. IEEE Transactions on Aerospace and Electronic Systems, AES-14(3):487--493, May 1978.

....rotation are ignored In this subsection we elaborate on the importance of using the AFTR even for an algorithm which calculates depth based on the shifts alone. Ignoring the AFTR amounts to taking it to be a unity matrix. This question has been investigated extensively by Mostafavi and Smith in [27, 28]. Their results are summarize below. For images having a circularly symmetric Gaussian correlation function, R( u ; v ) exp ae Gamma 1 2 Delta 2 [ 2 u 2 v ] oe ; 11) where u ; v are the spatial shifts, and Delta the correlation width , the effects of noncompensated ....

Mostafavi H. and F.W. Smith. Image correlation with geometric distortion, ii: Effect on local accuracy. IEEE Trans. on Aerospace and Electronic Systems, 14(3):494--500, May 1978.

....rotation are ignored In this subsection we elaborate on the importance of using the AFTR even for an algorithm which calculates depth based on the shifts alone. Ignoring the AFTR amounts to taking it to be a unity matrix. This question has been investigated extensively by Mostafavi and Smith in [27, 28]. Their results are summarize below. For images having a circularly symmetric Gaussian correlation function, R( u ; v ) exp ae Gamma 1 2 Delta 2 [ 2 u 2 v ] oe ; 11) where u ; v are the spatial shifts, and Delta the correlation width , the effects of noncompensated ....

....to provide a high enough probability of detecting the correct correlation peak and a low enough probability of locking onto a wrong peak; this is equivalent to 15 dB in power ratio. Taking a window of size 21 Theta 21, Delta of 1.5 pixels is 0:07 of the window size. Using the equations in [27]) we get the capture zone for a (or b) as Sigma1:32 pixels. We similarly calculated the capture zone for the expansion factor as: s = 0:871 to 1.148, and for the rotation as: Sigma8:5 0 . We have thus shown that the capture zone is quite wide. Images from real scenes are highly ....

[Article contains additional citation context not shown here]

Mostafavi H. and F.W. Smith. Image correlation with geometric distortion, part i: Acquisition performance. IEEE Trans. on Aerospace and Electronic Systems, 14(3):487-- 493, May 1978.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC