G. Borgefors, "Distance transforms in digital images, " Computer Vision, Graphics and Image Processing, vol. 34, pp. 344--371, 1986.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....To overcome the problem, we need an edge indicator function invariant to object motion. We choose the inverted distance transform of the edge map [17] 18] 4) where is the Euclidean distance from to , and is an arbitrary constant such that everywhere. An algorithm for computing is given in [19]. Threedimensional views of this function for some simple edge maps are shown in Fig. 4. Since the inverted distance transform has a constant gradient almost everywhere, it can guide a contour moving to edges over a long distance. In order to increase tracking stability, should be close to the ....

G. Borgefors, "Distance transforms in digital images," Comput. Vis., Graph., Image Process., vol. 34, pp. 344--371, 1986.

....edge detector is unbiased and identically normally distributed with variance # 2 , finally we obtain: p(x # # #) # p(c(x) x # #) # exp[ 1 # 2 D(x, #) 2 ] 2) where D(x, #) is the distance from x to the edge map. This distance can be computed by using the algorithm developed in [3]. Considering (2) the first requirement means that the expected contour should be composed of pixels which are as closest as possible to edges. Note that we have ignored in p(x # # #) the influence of the edge strength which in turn is related to the gradient, due to the drawbacks mentioned in ....

G. Borgefors. Distance transforms in digital images. Comp. Vision, Graph. and Image Proc., 34:344--371, 1986.

....independent (Fig. 3b) Moreover, objects with irregular shapes produce to many markers (Fig. 3f) because the geometric interpretation is drawn from one single level. Figure 3c and 3g show the separation results obtained by a reconstruction that starts with the maxima of the distance transform [24]. Although the distance transform is able to handle objects with different size (Fig. 3c) the irregular shape of cells again causes oversegmentation (Fig. 3g) However, progressive segmentation using markers and applying complete object information afterwards is a common concept in mathematical ....

G. Borgefors. Distance transforms in digital images. Computer Vision, Graphics and Image Processing, 34:679--698, 1986.

....points Q, and it has also been called the Voronoi distance of Q by analogy to Voronoi diagrams, which specify the locations equidistant from two or more points of a given set. 4. 1 The Voronoi distance D q (i; j) There are many methods of computing the Voronoi distance or an approximation of it [3], 5] We use an algorithm similar to [12] which produces distance transform values that are exact up to machine precision for some norm of the distance. This algorithm first processes each row independently. For each row of Q(i; j) it calculates the distance to the nearest pixel in the row, ....

Borgefors G. "Distance transforms in digital images", ieee Transactions on pattern analysis and machine intelligence, Vol. 8, 1986, p. 344371.

....result, so that computation times may become quite high. The second definition used for a skeleton is that of the ridge lines formed by the centers of all maximal disks included in the original shape, connected to preserve connectivity. This leads directly to the use of distance transforms [4], which can be computed in only two passes on the image. However, it is difficult to compute the distance transform without storing the whole image in memory. In our group, we have been testing both approaches. For thinning, we applied the well known algorithm illustrated by Fig. 1. This must be ....

G. Borgefors. Distance Transforms in Digital Images. Computer Vision, Graphics and Image Processing, 34:344--371, 1986.

.... a skeleton and a width value at each point on the skeleton (the so called quench function) For polygonal contours, the medial axis and the quench function can be computed in time linear in the number of vertices [CSW95] For pixel chain contours, this can be computed using the distance transform [Bor86] The dissimilarity of contours can be based on sample points along the contour curve, the whole contour curve, or the enclosed area. For example, Fourier descriptors are based on samples of the contour. A number of methods based on the contour curve and the area are mentioned below. 6.1 Turning ....

G. Borgefors. Distance transforms in digital images. Computer Vision, Graphics, and Image Processing, 34:344-371, 1986.

....over and over again. This turns out to be very well suited to our application as we generally want to compare one particular unidentified character to each of several plausible model characters. There are at least a few good ways to compute the distance transform in an efficient manner. The first [2] uses several serial passes of the data to propagate out estimations 6 of distance from ON pixels, where the distance is 0. Another method [11] makes separate vertical and horizontal sweeps of the data to deduce its distances. One reason that we decided upon the latter algorithm is that it is ....

....they are so far away that these comparisons can be eliminated. Since we are only interested in examining matches that are very close in nature, we can choose a very low threshold. This low threshold value can greatly increase the efficiency of the algorithm in [11] whereas the other algorithm [2] is not as well suited to this particular constraint. We refer the reader to the references for details of the algorithms. 6 The Learning Phase: Details The main goal for the learning phase is to develop a set of model characters for a corpus 7 . While we want to keep this phase as automated as ....

G. Borgefors. "Distance transforms in digital images." IEEE Trans. Pat. Anal. an Mach. Intel., 34:344--271, 1986.

.... transform of B[k; l] That is, the array D 0 [x; y] is zero wherever A[k; l] is one, and the other locations of D 0 [x; y] specify the distance to the nearest nonzero point of A[k; l] There are a number of methods for computing the rasterized Voronoi surface, or distance transform (e.g. [5, 15]) which we discuss briefly below. Proceeding with the analogy to the continuous case, we can compute the pointwise maximum of all the translated D and D 0 arrays to determine the Hausdorff distance as a function of translation (only now we are limited by the rasterization accuracy of the ....

.... section we summarize some of the approaches that we have used for computing the distance transform D[x; y] of a binary array E[x; y] where we denote the nonzero pixels of E[x; y] by the point set E) One method of computing D[x; y] is to use a local distance transform algorithm such as that in [5], 10] or [15] In practice we use a two pass serial algorithm that approximates the distance transform using a local mask to propagate distance values through the array (such as that of [5] Better distance values can be obtained using a method such as that of [15] which produces distance ....

[Article contains additional citation context not shown here]

G. Borgefors. Distance transforms in digital images. IEEE Trans. Pat. Anal. and Mach. Intel., 34:344--371, 1986.

No context found.

G. Borgefors, "Distance transforms in digital images, " Computer Vision, Graphics and Image Processing, vol. 34, pp. 344--371, 1986.

No context found.

G. Borgefors, "Distance transforms in digital images," Comp. Vision, Graph. and Image Proc., vol. 34, pp. 344--371, 1986.

No context found.

G. Borgefors, "Distance transforms in digital images," Comp. Vision, Graph. and Image Proc., vol. 34, pp. 344--371, 1986.

No context found.

G. Borgefors, "Distance Transforms in Digital Images," Computer Vision, Graphics, and Image Processing, vol. 34, pp. 344-371, 1986.

No context found.

G. Borgefors. Distance Transforms in Digital Images. Computer Vision, Graphics and Image Processing, 34:344--371, 1986.

No context found.

G. Borgefors, "Distance transforms in digital images," Computer Vision, Graphics and Image Processing 34, pp. 679--698, 1986.

No context found.

G. Borgefors. Distance transforms in digital images. Computer Vision, Graphics and Image Processing, 34:344--371, 1986.

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