Q. Chen, M. Defrise, and F. Deconinck. Symmetric phase-only matched filtering of Fourier-Mellin tranforms for image registration and recognition. IEEE trans. on Pattern Analysis and Machine Intelligence, 16(12):1156--1168, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... can be recovered by applying the Fourier transform and using phase correlation [2] A Fourier based method to accommodate translation and rotation was described in [3] The Fourier Mellin transform has been introduced to register images that are misaligned due to translation, rotation, and scale [4, 5, 6, 7]. This method applies a Fourier transform to images to recover translation. Then a log polar transformation is applied to the magnitude spectrum and the rotation and scale is recovered by using phase correlation in the log polar space. This method exploits the fact that by operating on the ....

....to recover the rotation angle and scale factor between the pair of input images. The problem here, though, is that limited scale factors can be determined because large scale factors would alter the frequency content beyond recognition. It should be noted that the maximum scale factor recovered in [6] and [7] is 2.0 and 1.8, respectively. The work presented in this paper consists of two modules: log polar registration followed by optimization based affine registration. The latter module is similar in spirit to the method described in [8] However, since it is based on optimization techniques, ....

Q. Chen, M. Defrise, and F. Deconinck. Symmetric phaseonly matched filtering of fourier-mellin transforms for image registration and recognition. IEEE Trans. Pattern Analysis and Machine Intelligence, 16(12):1156--1168, December 1994.

....the image can be 4 scaled. Instead of rescaling the image, scale invariant segmentation can be obtained by, e.g. including image patches at different scales in the training set [15] Another approach entails transforming the image by an invariant mapping such as the Fourier Mellin transform [16, 17]. Scale invariant segmentation can also be obtained by training a classifier based on features that eliminate changes in scale. Such scale invariant features include wavelets [5] features from the linear scale space [18] and different statistical moments [19, 20] Statistical moments have the ....

Q. Chen, M. Defrise, F. Deconinck, Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition, IEEE Transactions on Pattern Analysis and Machine Intelligence 16 (12) (1994) 1156-1168.

....require each entry of the index to be of length 255 for grey images or 3 255 for colour images. For example, if two photos were taken of Billy Idol, one in the day and one in the night, the colours would be completely di#erent even though they contain the same person. A paper by Chen et al. CDD94] gives details of the Fourier Mellin Invariant Descriptor (FMID) matching method. The magnitude of the Fourier transform of a signal is translational invariant. The polar log transform of a signal converts any rotations and scalings into translations. Therefore to remove any translations, ....

....[Hay99] The data was taken from the TREC database [ST01] and from results given by search engines on the Web. The results showed that FDS was superior in both cases. This is due to the fact that the cosine measure is a specialised case of FDS. Interesting ideas were obtained from a paper by [CDD94] on methods to compare images to find similarities. The Fourier Mellin Invariant Descriptor method uses certain characteristics of the Fourier transform to remove any inconsistencies due to translation, scaling and rotations of the images. Other similar methods are the Analytical Fourier Mellin ....

Qin-Sheng Chen, Michel Defrise, and F. Deconinck. Symmetric phase-only matched filtering of fourier-mellin transforms for image registration and recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 16(12):1156--1168, December 1994.

....correlation for less correlated images and hence would be inappropriate for detecting similarity. 3. TRANSLATION AND ROTATION INVARIANT DESCRIPTION As is well known, an image and its translated, rotated and scaled version are closely related after application of the Fourier Mellin transform [1], 10] We modify this slightly to get a translation and rotation invariant transform. Let f1 (x; y) be the original image, and f2 (x; y) its translated and rotated version. Then f2 (x; y) f1(xcos 0 ysin 0 Gamma x0 ; Gammaxsin 0 ycos 0 Gamma y0 ) 4) where (x0 ; y0) is the translation ....

....characteristics quite well, in both spatial and Fourier amplitude domains. To this capability, we add the phase information from the WPT and the corresponding reduction in data provided by the WPT multiresolution. While the attributes of phase and phase only filters has been well recognized [7] [1], our experiments confirm the phase discrimination power and robustness. As demonstrated, the multiresolution phase maintains the discrimination capability quite well upto the 32 Theta 32 size scale. At the same time, it also preserves the local edges well enough, thereby making the algorithm ....

[Article contains additional citation context not shown here]

Q.S.Chen, M.Defrise and F.Deconinck, "Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition", IEEE Trans. Pattern Anal. Machine Intell., vol.16, no.12, pp.1156-1168, 1994.

....scales of resolution. Examples of this are scale space, Gabor and wavelet decomposition methods. The multiresolution methods deserve particular attention because of their relation with the human visual system. Psychovisual research offers evidence that humans process images in a multiscale way [10]. The early processing in the brain performs a kind of spatial frequency analysis, and consequently, the visual cortex has separate cells that respond to different frequencies and orientations [11] This multiscale processing, which humans obviously apply very successfully to texture perception, ....

....and wavelet packet (right) decompositions (d = 2) 18 CHAPTER 1. METHODS FOR TEXTURE ANALYSIS 1.3. 3 Application Examples In the past multiresolution texture analysis techniques have led to very good results [28] 29] 30] This has been the case with the use of Gabor filters [45] 46] 47] 48] [10] [49] In the last few years, comparable results are obtained with the use of wavelet transforms. 50] 23] 20] 24] 51] 52] 25] The wavelet texture analysis is rapidly finding its way to real world applications. In a growing number of areas, the methods are being investigated. An online ....

[Article contains additional citation context not shown here]

Q.S. Chen, M. Defrise, and F. Deconinck. Symmetric phase only matched filtering of fourier mellin transforms for image registration and recognition. IEEE Trans. PAMI, 16(12):1156--1168, 1994.

....metric value, and applying the mapping to move each image pixel from its current position to the corrected position. A wide range of registration techniques has been developed for many different types of applications and data, such as mean squared alignment [42] Fourier descriptor matching [13], correlation registration [43] moment invariant matching [18] maximizing mutual information [17, 103] and others. Basically, each method exploits different image features of the object such as boundaries, moments or texture, which are extracted from 47 the image, and uses them to solve the ....

Q. Chen, M. Defrise, and F. Deconinck, "Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 16, no. 12, pp. 1156--1168, Dec 1994.

.... can be recovered by applying the Fourier transform and using phase correlation [2] A Fourier based method to accommodate translation and rotation was described in [3] The Fourier Mellin transform has been introduced to register images that are misaligned due to translation, rotation, and scale [4, 5, 6, 7]. This method applies a Fourier transform to images to recover translation. Then a log polar transformation is applied to the magnitude spectrum and the rotation and scale is recovered by using phase correlation in the log polar space. This method exploits the fact that by operating on the ....

....to recover the rotation angle and scale factor between the pair of input images. The problem here, though, is that limited scale factors can be determined because large scale factors would alter the frequency content beyond recognition. It should be noted that the maximum scale factor recovered in [6] and [7] is 2.0 and 1.8, respectively. The work presented in this paper consists of two modules: log polar registration followed by optimization based affine registration. The latter module is similar in spirit to the method described in [8] However, since it is based on optimization techniques, ....

Q. Chen, M. Defrise, and F. Deconinck. Symmetric phaseonly matched filtering of fourier-mellin transforms for image registration and recognition. IEEE Trans. Pattern Analysis and Machine Intelligence, 16(12):1156--1168, December 1994.

....in the brain to single trial visual stimulus. Pattern recognition of the stimulus is performed in the phase domain. The phase of the signal contains almost all the information about the localisation of events on the time axis . The use of phase for pattern recognition is supported by Chen et al. [5]. Their work presents a new method of matching a 2 dimensional image to a translated, scaled and rotated reference image. They use Fourier Mellin descriptors for the images that are matched using Symmetric Phase Only Match Filtering (SPOMF) Phase only match filtering is used as the spectral ....

Q. Chen, M. Defrise and F. Decconinck, "Symmetric phase-only matched filtering of Fourier-Mellin Transforms for image registration and recognition", IEEE Trans. PAMI, 16(12), pp.1156-1168, 1994.

....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 ....

Q.-S. Chen, M. Defrise, and F. Deconinck. Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 16(12):1156-- 1168, December 1994.

....with all those of database objects would require high computational cost and estimation of j may be inaccurate if the correlation curve shows a broad peak. An alternative fast and accurate method is to use the Fourier magnitude spectrum for similarity measure and the Fourier phase for estimating j [4]. Let f be a chromaticity distribution function in the new coordinate system . The deformation by G results in the translation of f , i.e. f G ( f( j) and their Fourier transforms will differ only in their phases: F [f G ( e Gammai j; F [f( 10) Note that the ....

Q.S. Chen, M. Defrise, and F. Deconinck. Symmetric phase-only matched filtering of fourier-mellin transforms for image registration and recognition. IEEE Trans. PAMI, 16:1156--1168, 1994.

....Our preliminary results [3] have shown good discrimination of patterns and an insensitivity to the effects of finite resolution. This article presents a detailed derivation for the procedure. Our algorithm addresses the problem of evaluating the similarity between objects in two dimensional images [9]. Given a gray level reference image and a gray level input image, the algorithm aims to identify if an object in the input image corresponds to the pattern object in the reference image, regardless of scaling factor, rotation angle, and translation. We can divide image recognition techniques into ....

....the object shape and the phase spectrum contains information about the object translation; therefore, the shape information in the magnitude spectrum is naturally translation invariant. The Fourier Mellin transform is a powerful tool for image recognition techniques that use this spectral property [12, 9]. This method represents rotation and scaling as single translations on the parameter space, and allows the use of the Phase Correlation Technique [10] to determine translation, scaling and rotation. Thus, there are four free parameters. To represent scaling and rotation as translations, the ....

[Article contains additional citation context not shown here]

M. Defrise Q. Chen and F. Deconinck. Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition. IEEE Trans. Pattern Anal. Mach. Intel., 16(12):1156--1168, December 1994.

No context found.

Q. Chen, M. Defrise, and F. Deconinck. Symmetric phase-only matched filtering of Fourier-Mellin tranforms for image registration and recognition. IEEE trans. on Pattern Analysis and Machine Intelligence, 16(12):1156--1168, 1994.

No context found.

Q. Chen, M. Defrise, and F. Deconinck, "Symmetric phase-only matched filtering of fourier-mellin transforms for image registration and recognition," IEEE Trans. Pattern. Anal. Machine Intell., vol. 16, no. 12, pp. 1156--1168, Dec. 1994.

No context found.

Chen, Q.; Defrise, M.; Deconinck, F. Symmetric phase-only matched filtering of FourierMellin transforms for image registration and recognition. IEEE Trans. Pattern Analysis and Machine Intelligence 16:1156-1168; 1994.

No context found.

Chen, Q.-s.; Defrise, M.; Deconinck, F. Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition. IEEE Trans. Pattern Analysis and Machine Intelligence 16: 1156-1168; 1994.

No context found.

Chen, Q.-S., Defrise, M., & F. Deconinck, "Symmetric phase-only matched filtering of Fourier-Mellin transforms for image registration and recognition," IEEE Trans. Pattern Anal. Mach. Intell. 16, 1156-1168 (1994).

No context found.

Q. Chen, M. Defrise and F. Decorninck, "Symmetric phase-only matched filtering of FourierMellin transforms for image registration and recognition," IEEE Trans. on Pattern Recognition and Machine Intelligence, 12(12), 1156-1198, (1994).

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