E. D. Castro and C. Morandi. Registration of translated and rotated images using finite fourier transforms. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI9 (5):700--703, Sept. 1987.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... By computing the phase correlation via the Fast Fourier Transform, robust results are achieved, which can be implemented on real time hardware [16] Faster computation can be achieved at the cost of robustness and accuracy if the phase correlation is computed only with one dimensional FFTs [1] [7] and [12] propose frequency domain algorithms that can search for rotation and affine transformations respectively, but they are computationally expensive. Figure 2: Mapping to a 2 D World Model However, the image rotation (about the optical axis) and scaling information of an image can be found ....

E. De Castro and C. Morandi, "Registration of Translated and Rotated Images Using Finite Fourier Transforms," IEEE Transactions on PAMI, Vol. 9, No. 5, September 1987, pp. 700703.

....level. The traditional approach in the field typically defines pattern matching using concepts from continuous mathematics and uses fast Fourier transform to implement translations (as a convolution) and polar transformations (or sampling on an evenly spaced grid) to implement rotations; see e.g. [2]. These methods are inherently approximate and hence introduce another layer of inaccuracy on top of the noise in the data that can be very high in typical applications. Therefore we believe that our exact combinatorial analysis and methods can be useful. We are not aware of earlier similar ....

E. D. Castro and C. Morandi. Registration of translated and rotated images using finite fourier transforms. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI9 (5):700--703, Sept. 1987.

....strings. A work supported by ComBi and the Academy of Finland under grant 44449. y The corresponding author. Previous work on image translations and rotations has concentrated on methods that typically use template matching, cross correlation (e.g. Caelli and Liu, 1988) and FFT techniques (Castro and Morandi, 1987). For a review, see e.g. Wood, 1996) Combinatorial techniques that resemble ours were used in (Landau and Vishkin, 1994) and in (Takaoka, 1996) Our method is a generalization of our work of exact matching reported in (Fredriksson and Ukkonen, 1998) 2 Problem definition Let pattern P = P ....

Castro, E. D. and Morandi, C. (1987). Registration of translated and rotated images using finite fourier transforms. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-9(5):700--703.

....the images are bandlimited and accordingly Nyquist sampled, and periodic with the image size. Images are often preprocessed in order to make these assumptions more valid. For example, Gaussian smoothing can be applied to limit the bandwidth. In an extension of the phase correlation technique, De Castro 87] has proposed a technique to register images which are both translated and rotated with respect to each other. Rotational movement, by itself without translation, can be deduced in a similar manner as translation using phase correlation by representing the rotation as a translational displacement ....

....Rotational movement, by itself without translation, can be deduced in a similar manner as translation using phase correlation by representing the rotation as a translational displacement with polar coordinates. But rotation and translation together represent a more complicated transformation. De Castro 87] present the following two step process to first determined the angle of rotation and then determine the translational shift. Rotation is invariant with the Fourier Transform. Rotating an image, rotates the Fourier transform of that image by the same angle. Two images I 1 (x; y) and I 2 (x; y) ....

[Article contains additional citation context not shown here]

E. De Castro and C. Morandi "Registration of Translated and Rotated Images using Finite Fourier Transforms," IEEE Trans. on Pattern Analysis and Machine Intelligence PAMI-9 ,No. 5, September 1987, pp700-703.

.... (k; l) are related to the world coordinate frame by: k = a 11 x a 12 y a 13 z a 14 (6) l = a 21 x a 22 y a 23 z a 24 (7) Assuming that camera motions from one low resolution frame to another are only composed of translation and rotation around the origin of the world frame [8]: a 11 = a 22 = S x cos (8) Gammaa 12 = a 21 = S y sin (9) where S x and S y are the ratios of the sampling rates of the desired high resolution images to those of the low resolution frames along x and y axes, respectively, and is the rotation angle of the camera around the origin of the ....

....is a 2 Theta 4 matrix containing camera parameters. The last column of A contains the interframe shift parameters, specifying the coregistration of image frames. Other parameters specify the sampling rate along the two axes and the rotation of the camera around the origin (for further details see [8][20] RR n2706 18 H. Shekarforoush, M. Berthod, J. Zerubia ....

E. De Castro and C. Morandi. Registration of translated and rotated images using finite fourier transforms. In IEEE PAMI, volume 9, No 5, pages 700--703, 1987.

....in another image. To find the optimal match for the template in the image, the proper alignment between the image and template must be found. A transformation must be found so that the points in one image can be related to their corresponding points in the other. Definition Image registration[28] can be defined as a mapping between two images both spatially and with respect to intensity. If these images are expressed as two 2D arrays of a given size denoted by I 1 and I 2 , where I 1 (x; y) and I 2 (x; y) each map to their respective intensity values, then the mapping between images can ....

....noise, which is spread across all frequencies, then the location of the peak will be inaccurate since the phase difference at each frequency is corrupted. In this case, finding the peak of spatial cross correlation is good. In an extension to the phase correlation technique, De Castro and Morandi [28] have proposed a technique to register images which are both translated and rotated with respect to each other. Rotational movement can be deduced in a similar manner as translation using phase correlation by representing the rotation as a translation displacement of polar coordinates. But ....

E. Castro, and C. Morandi, "Registration of translated and rotated images using finite Fourier Transforms", IEEE Trans. Pattern Analysis and Machine Intelligence, V.9, #5, pp. 700-703, 1987.

....of estimating this offset at subpixel accuracy, and using this estimate to register the second image to the grid of the first image. This problem has been attacked in the signal or pixel domain [1, 6, 7, 9, 10, 11, 14, 15, 16, 17, 18, 19, 20, 22, 24, 23, 27, 30, 31, 33] and in the Fourier domain [2, 3, 4, 5, 7, 8, 12, 13, 21, 25, 26, 28, 29, 32]. Our work follows the latter body, and estimates the shift from basic phase relationships between the Fourier transform of the two images. However, unlike previous work, we do not assume that each observed image represents alias free samples of an underlying continuous image. In fact, point ....

E. De Castro and C. Morandi. Registration of translated and rotated images using finite Fourier transforms. IEEE Trans. on Pattern Analysis and Machine Intelligence, PAMI9 (5):700--703, 1987.

....developed before 1993, one is referred to reference [1, 2] A recent comparative study of some of these methods can be found in [3] Generally speaking, rigid body registration of images with similar contrast is a solved problem. Many methods, including the Fourier transform based method [4, 5] or the voxel similaritybased method [6 8] can align images of this type with sub pixel accuracy. However, for images with significant contrast differences, such as those from different imaging modalities, the problem has been challenging because of the lack of an effective matching criterion. To ....

E. De Castro and C. Morandi "Registration of translated and rotated images using finite Fourier transforms," IEEE Trans. PAMI, vol. 9, pp. 700-703, 1987.

....sensed image(s) is to be established. Common diculty here is that there could be some landmarks having no corresponding counterparts in the other frame(s) Matching methods can be based on parameter clustering [30] graph matching [14] relaxation [26] 33] correlation like techniques [25] 2] [4] taking into account the landmark neighborhood or hybrid invariant approaches [34] 10] 35] 8] 6] This work has been supported by the grant No. 624178 3 of the Grant Agency of the Czech Ministry of Health. The authors thank to Dr. Igor Hollander for providing the CT data. Estimation ....

E. De Castro and C. Morandi. Registration of translated and rotated images using nite Fourier transforms. IEEE Trans. Pattern Analysis and Machine Intelligence, 5:700-703, 1987.

....Within these, the use of Fourier techniques has received considerable attention, with the symmetry and invariant properties of the Fourier domain being exploited to derive affine estimates between images and image regions. Examples include methods based on moments [3] generalised correlation [4] [6] and other schemes [7] 8] The method described here is also a frequency domain technique. It assumes a piecewise affine transformation field and uses a generalised correlation scheme to estimate the local transformations. The underlying principle is the same as that in [4] 6] transform to ....

....correlation [4] 6] and other schemes [7] 8] The method described here is also a frequency domain technique. It assumes a piecewise affine transformation field and uses a generalised correlation scheme to estimate the local transformations. The underlying principle is the same as that in [4] [6] transform to the frequency domain and align the spectra prior to correlating. The required spectral alignment corresponds to the linear component of the affine transformation and can be estimated by relating the shift invariant magnitude spectra. An estimate of the translation component then ....

[Article contains additional citation context not shown here]

E.De Castro and C.Morandi, "Registration of translated and rotated images using finite Fourier transforms", IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 9(5), pp. 700-703, 1987. British Machine Vision Conference 325

.... (of original images or extracted feature images) 203, 204, 205, 206, 207, 208, 209, 210, 196, 211, 212, 213, 214, 215, 49, 216, 217, 197, 218, 132, 131, 219, 220, 221, 7, 55, 222, 223, 224, 225, 200, 226, 227, 228, 229] # Fourier domain based cross correlation, and phase only correlation [230, 231, 232, 228, 233, 234]. # Minimization of variance of intensity ratios [207, 90, 235, 236, 224, 225, 237] # Minimization of variance of grey values within segments [238, 130] # Minimization of the histogram entropy of difference images [239] # Histogram clustering and minimization of histogram dispersion [207, ....

E. de Castro and C. Morandi. Registration of translated and rotated images using finite Fourier transforms. IEEE Transactions on pattern analysis and machine intelligence, 5:700--703, 1987.

....4. Phase Correlation Based Registration In phase correlation based registration, it is the 2D Fourier Transform of an image, rather than the image itself, which is processed, in the search for the transformation of the distorted image Irish Signal and Systems Conference, Galway, June 1999 [3]. This is motivated by the fact that certain geometric distortions (translation, rotation and scaling) are mapped into a more computationally tractable form in the frequency domain. The computational efficiency of the FFT, O(nlog 2 n) also makes this method attractive. It is based on the ....

....the correct mappings are represented by a tight cluster of data points. The task therefore in vector mapping is to accurately locate this cluster in the data set. An algorithm which splits the data set up into 20202020 blocks and counts the number of data points in each one has been implemented [3]. The blocks with the largest amount of objects in them, were further split, and the block resulting with the largest amount of objects, was taken to be the largest cluster. The centre of this block was taken as the best estimate of the transform coefficients. Alternative methods for locating the ....

E. De Castro, and C. Morandi, "Registration of Translated and Rotated Images Using IEEE Trans Patt. Anal. Mach. Intell., vol. 9, no. 5, pp. 700--703.

....of this transformation are then estimated by optimizing a similarity criterion between one image and the other transformed image. The geometric transform can be rigid or non rigid. Examples of rigid transformations can be found in [37, 152, 194, 200] which include only translation, and [38, 98], which include both translation and rotation. Translation and rotation are the elementary components of a rigid body transformation. In addition to rigid transformation, geometrical scaling is sometimes necessary for photographic views based on the use of a lens system because there is a direct ....

....proposed including sign change criteria [193] The similarity criterion can be computed for every possible value of the registration parameters. This is rather computationally expensive. One group has used phase correlation 24 Fourier methods to estimate translation and rotation parameters [38]. Another group, which searched a space of bilinear geometric transforms to improve digital subtraction angiographic images of the coronary arteries, used genetic algorithms and simulated annealing to reduce the search time and the problem of local optima [137] A more recent solution uses a ....

E. De Castro and C. Morandi. Registration of translated and rotated images using finite Fourier transforms. IEEE Trans. Pattern Anal. Mach. Intell., 9:700--703, 1987.

....more robust than the feature based method in some situations. However, if the orientation difference between the two images is large, the value of cross correlation will be greatly influenced and the correspondences between feature points, thus, hard to derive. Therefore, De Castro and Morandi [13] proposed an elegant method, called phase correlation, to overcome this problem. However, when the overlapping area between images is small, their method becomes unreliable. In order to solve the problem, it is necessary to develop a method to estimate the rotation parameter in advance. In [3] ....

E. De Castro and C. Morandi, "Registration of translated and rotated image using finite Fourier transform," IEEE Trans. Pattern Anal. Machine Intell., vol. 9, no. 5, pp. 700--703, Sep. 1987.

....to shift from one map to the next. These control points are found from pairs of edges common to both maps (Janssen Vossepoel [2] Dani Chaudhuri [3] used a normalized correlation method with a multiresolution pyramid for unsupervised assembly of montages of greyscale images. Castro Morandi [4] used Fourier transforms to speed up the registration of images. Although this is useful for template matching, the process requires human intervention to mask out regions not common to both images when used for image mosaicing. 2. Montage assembly using XOR correlation The cross correlation c(x; ....

Castro, E. de and Morandi, C., "Registration of Translated and Rotated Images using Fourier Transforms," IEEE PAMI, 9, 5, September, 1987, pp700--703.

....Within these, the use of Fourier techniques has received considerable attention, with the symmetry and invariant properties of the Fourier domain being exploited to derive affine estimates between images and image regions. Examples include methods based on moments [3] generalised correlation [4] [6] and other schemes [7] 8] The method described here is also a frequency domain technique. It assumes a piecewise affine transformation field and uses a generalised correlation scheme to estimate the local transformations. The underlying principle is the same as that in [4] 6] transform to ....

....correlation [4] 6] and other schemes [7] 8] The method described here is also a frequency domain technique. It assumes a piecewise affine transformation field and uses a generalised correlation scheme to estimate the local transformations. The underlying principle is the same as that in [4] [6] transform to the frequency domain and align the spectra prior to correlating. The required spectral alignment corresponds to the linear component of the affine transformation and can be estimated by relating the shift invariant magnitude spectra. An estimate of the translation component then ....

[Article contains additional citation context not shown here]

E.De Castro and C.Morandi, "Registration of translated and rotated images using finite Fourier transforms", IEEE Trans. Pattern Analysis and Machine Intelligence, vol. 9(5), pp. 700-703, 1987.

....transformation directly by search, this has proved computationally expensive [30] Most approaches in use attempts to solve the motion equations directly by adopting some feature based scheme [32, 33] which can involve the solution of as many as 15 simultaneous non linear equations. De Castro [34] used Fourier methods to estimate rotation. Adelson et al. 29] use a coarse to fine gradient based motion estimation scheme to determine optical flow. An affine model is then fitted to the found optical flow by means of some least squares fitting procedure. The affine model and estimation scheme ....

E. De Castro and C. Morandi, "Registration of Translated and Rotated Images Using Finite Fourier Transforms," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 9, pp. 700--703, September 1987.

.... an affine transformation in frequency, which will be discussed in detail in the next chapter [21, 81] It was noted in 1987 by De Castro and Morandi that rotation can be estimated by polardomain correlation, exploiting the fact that the Fourier transform is rotation invariant in the polar domain [37]. However, if the rotation is also combined with a translation, the method degenerates it is computationally demanding, requiring up to 180 inverse Fourier transforms. This approach was improved computationally and extended to handle dilations by noting analogously that the Fourier transform is ....

E. De Castro and C. Morandi. Registration of translated and rotated images using finite Fourier transforms. IEEE Transactions on Pattern Analysis and Machine Intelligence, 9(5):700--703, September 1987.

No context found.

E. D. Castro and C. Morandi. Registration of translated and rotated images using finite fourier transforms. IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI9 (5):700--703, Sept. 1987.

No context found.

E. D. Castro and C. Morandi, Registration of translated and rotated images using nite Fourier transforms, IEEE Transactions on Pattern Analaysis and Machine Intelligence 9(5), pp. 700703, 1987.

No context found.

E. De Castro and C. Morandi, "Registration of translated and rotated images using finite Fourier transforms," IEEE Trans. Pattern Analysis and Machine Intelligence 9, pp. 700--703, 1987.

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