R.D. Morris. Image Sequence Restoration using Gibbs Distributions. PhD thesis, Cambridge University, England, 1995.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....Kalman filter. A new Bayesian restoration technique, dealing with both low and high frequencies around and inside the detected artifacts, is investigated to achieve a nearby invisible restoration of damaged areas. 1 Introduction Digital film restoration has only very recently been explored [11] [13][16] 14] 5] 2] 4] 9] Up to now, old films have been restored using traditional film restoration techniques, like chemical baths or film polishing. However, such techniques do not permit the removal of all kinds of degradations. Computer aided techniques have also been used to restore classic old ....

....algorithms must also preserve the visual quality of the films. # The processing power has to be cheap, processing time as short as possible and the restoration process should be automated, to cut exploitation costs. Most existing work on old film restoration is based on the video format [11][13][16] 14] 5] 4] If we apply these techniques on 2K images, they will involve high computational costs. Besides, motion picture (35mm) images require a much better visual quality than video, thus increasing the complexity of the digital restoration task. Our digital restoration system handles the ....

[Article contains additional citation context not shown here]

Morris R.D. Image sequence restoration using Gibbs distributions. PhD thesis, University of Cambridge, May 1995.

....model for the luminance variation to catch a local interaction of some neighbor pixels. One can then develop a method which fits this form to the observed image. The main interpolation models used in signal and image processing are : # polynomial models [7] 13] # stochastic models [14] [9], 8] 6] and # Fourier series [3] 1] Polynomial interpolation usually relies on splines or Bsplines representations which are the basis for curve fitting [7] The only requirement of such methods is the order of the polynomial which best fits the input signal. This is a good solution for ....

....for the reconstruction of homogeneous regions. However, it fails to reconstruct regions of high activity such as textured areas because it approximates only the low frequencies of the input signal. Stochastic models such as AutoRegressive (AR) models [6] or Markov Random Field (MRF) models [8] [9] are also widely used for spatial interpolation process. They allow a better interpolation than polynomial models. Indeed, the high frequencies which are lost in polynomial interpolation may be recovered using such stochastic interpolation procedure. However, they are difficult to implement. ....

Morris R.D. Image sequence restoration using Gibbs distributions. PhD thesis, University of Cambridge, May 1995.

....The sections immediately following describe the functional forms of the various probability distributions required and introduce a new prior for occlusion estimation. THE LIKELIHOOD : It is typical to associate the presence of an ON occlusion site with the violation of temporal image smoothness [5, 6], implicit in the image model proposed above. This would therefore involve the inclusion of an occlusion variable in the expression for the data likelihood. However, closer investigation of the behaviour of motion fields with respect to a moving object (see figure 1) indicates that occlusion can ....

R. D. Morris. Image Sequence Restoration using Gibbs Distributions. PhD thesis, Cambridge University, England, 1995.

....translation that optimises the correlation for a few horizontal and vertical lines. If the sensor is mounted in a moving vehicle or no apriori knowledge about the type of sensor motion is known, methods based on the model of a moving rigid planar patch [16] or optical flow techniques can be used [17]. Detection of moving targets Once the sensor motion is estimated, preceding images are warped onto the current one. Then the original image is subtracted from the warped ones. If a moving object is present in the scene, we should find a large value at its position. The resulting images after ....

R. Morris, "Image Sequence Restoration using Gibbs Distributions," in Dept. Of Engineering. Cambridge: Trinity College, 1995.

....region along its particular motion trajectory. Most of the work done mainly involves motion compensated temporal ltering techniques with appropriate 2D or 3D Wiener lter for noise suppression, 2D 3D median ltering or more appropri 2 ate morphological operators for removing impulsive noise [16, 38, 39, 31, 27, 52, 19, 17]. However, and due to the fact that image sequence restoration is an emerging domain compared to 2D image restoration, the literature is not so abundant than the one related to the problem of restoring just a single image. For example, numerous PDE based algorithms have been recently proposed for ....

....assumed to be static. A simple moving object detector can be obtained using a thresholding technique over the inter frame dioeerence between a so called reference image and the image being observed. Decisions can be taken independently point by point [73] More complex approaches can also be used [55, 57, 56, 1, 36, 45, 16, 38, 39, 31, 27, 52]. However, in our application, we are not just dealing with a motion segmentation problem neither just a restoration problem. In our case, the so called reference image is built at the same time while observing the image sequence. Also, the motion segmentation and the restoration are done in a ....

R.D. Morris. Image Sequence Restoration using Gibbs Distributions. PhD thesis, Cambridge University, England, 1995.

....represented by S n ( x) and ( v) is the weight associated with each clique. The situation is illustrated in the left hand portion of figure 1. Note that the cliques employed here assume first order interactions even though the eight connected neighborhood can involve some second order cliques [9]. In order to discourage smoothness over too large a range, v) is defined as ( v) Nw =j X( v) Gamma xj where X( v) is the location of the block providing the neighborhood vector v. This location is measured in terms of block lengths. As before, Nw is the number of uncorrupted pixels ....

....is measured in terms of block lengths. As before, Nw is the number of uncorrupted pixels in the block. Large ( v) encourages motion vector smoothness, and small oe 2 e encourages vectors which minimize the DFD. It is true that equations 2, 3, 4 are sufficient to estimate the motion field itself [9, 11], and the prior can be altered to account for motion discontinuities. However, a direct solution for the MAP estimate (2) with respect to the vector field (via some Monte Carlo technique) is computationally demanding [9] In practice, after the use of the AWM estimator and the blotch detector, it ....

[Article contains additional citation context not shown here]

R. D. Morris. Image Sequence Restoration using Gibbs Distributions. PhD thesis, Cambridge University, England, 1995.

....region along its particular motion trajectory. Most of the work done to date mainly involves motion compensated temporal ltering techniques with appropriate 2D or 3D Wiener lter for noise suppression, 2D 3D median ltering or more appropriate morphological operators for removing impulsive noise [4, 14, 15, 12, 9, 20, 6, 5]. However, and due to the fact that image sequence restoration is an emerging domain compared to 2D image restoration, the literature is not so abundant than the one related to the problem of restoring just a single image. For example, numerous PDE based algorithms have been recently proposed to ....

....using a thresholding technique over the inter frame dioeerence between a so called reference image and the image being observed. Decisions can be taken independently point by point [8] or over blocks in order to achieve robustness in noise inAEuence [25] More complex approaches can also be used [22, 1, 13, 19, 4, 14, 15, 12, 9, 20]. However, in our application, we are not just dealing with a motion segmentation problem neither just a restoration problem. In our case, the so called reference image is built at the same time while observing the image sequence. Also, the motion segmentation and the restoration are done in a ....

R. Morris. Image Sequence Restoration using Gibbs Distributions. PhD thesis, Cambridge University, England, 1995.

No context found.

R.D. Morris. Image Sequence Restoration using Gibbs Distributions. PhD thesis, Cambridge University, England, 1995.

No context found.

R. D. Morris, "Image sequence restoration using Gibbs distributions", Ph.D. thesis at University of Cambridge, UK, May 1995 7

No context found.

R. Morris, Image Sequence Restoration using Gibbs Distributions. PhD thesis, Dept. of Egineering and Trinity College, Cambridge University, 1995. 12

No context found.

R. Morris, Image Sequence Restoration using Gibbs Distributions. PhD thesis, Dept. of Egineering and Trinity College, Cambridge University, 1995.

No context found.

R. Morris, Image Sequence Restoration using Gibbs Distributions. PhD thesis, Dept. of Egineering and Trinity College, Cambridge University, 1995.

No context found.

R. Morris, Image Sequence Restoration using Gibbs Distributions. PhD thesis, Dept. of Egineering and Trinity College, Cambridge University, 1995.

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