Bhanu B. and Faugeras O., \Shape matching of two dimensional objects", PAMI 6, 137-155, 1984.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....such techniques, are not particularly well suited for IDB work. This is mostly due to the following reasons: most of them are model based , since they assume that the number and the kinds of objects (i.e. classes) to be recognized 20 are known in advance. Others, such as those proposed in [57, 58, 59, 60, 61, 53, 62, 63], perform an exhaustive search: all database objects or object class models are compared (one by one) against all query objects. To our knowledge, none of the above techniques has ever been tested on large databases storing hundreds or thousands of objects or models. An object recognition ....

Bir Bhanu and Olivier D. Faugeras. Shape Matching of Two-Dimensional Objects. IEEETransactions on Pattern Analysis and Machine Intelligence, 6(2):137--156, 1984.

.... and reconstruction [53,127,46,82,67,93] shape from shading [68] stereo, motion and optical ow [132,64,63,105,7] texture [61,73,30] edge detection [131,126] image segmentation [119,85] perceptual grouping [94,101,62] interpretation of line drawings [83] object matching and recognition [41,32,117,16,10,100,106,134,42,86,87], and pose estimation [58] In all of the above cited examples, the solution is explicitly de ned as an optimum of an objective function by which the goodness, or otherwise cost, of the solution is measured. Optimization may also be performed implicitly: the solution may optimize an objective ....

B. Bhanu and O. D. Faugeras. \Shape matching of two-dimensional objects". IEEE Transactions on Pattern Analysis and Machine Intelligence, 6(2):137{ 155, March1984.

.... which is represented as an ARG, two graph matching techniques are needed to relate the ARG of the object scene to a reference ARG (or a pattern represented as an ARG) Twograph matching as a fundamental problem in computer vision and pattern recognition research has been widely investigated [1,4,6,15,19,20,22,23]. However, little has been done for how to build an e#ective reference ARG. We advocate that the reference ARG should be learned from multiple sample 2 hong (a) b) Fig. 2. An ARG that represents a group of image segments. a) A group of image segments in di#erent colors, b) the ARG ....

Bhanu, B., and O. D. Faugeras, Shape Matching of Two-Dimensional Objects, IEEE Trans. Pattern Analysis and Machine Intelligence, 6 (1984), 137-156.

.... and reconstruction [53,127,46,82,67,93] shape from shading [68] stereo, motion and optical ow [132,64,63,105,7] texture [61,73,30] edge detection [131,126] image segmentation [119,85] perceptual grouping [94,101,62] interpretation of line drawings [83] object matching and recognition [41,32,117,16,10,100,106,134,42,86,87], and pose estimation [58] In all of the above cited examples, the solution is explicitly de ned as an optimum of an objective function by which the goodness, or otherwise cost, of the solution is measured. Optimization may also be performed implicitly: the solution may optimize an objective ....

B. Bhanu and O. D. Faugeras. \Shape matching of two-dimensional objects". IEEE Transactions on Pattern Analysis and Machine Intelligence, 6(2):137{ 155, March 1984.

....methods. From another view, image segmentation is a labeling problem; the ideal segmentation should be globally consistent or nearest to the one with maximum likelihood. This has led to various relaxation labeling or stochastic labeling methods that are related to general optimization algorithms [3, 17, 12]. Nearly all require some prior information, such as the number of labels needed or the probability distribution of labels in the image. Such information is not always available for general imagery. After defining the criterion function for labeling, the next problem is computing the solution to ....

B. Bhanu and O. Faugeras. Shape matching of twodimensional objects. PAMI, 6(2):137--156, 1984.

....Little or no emphasis is given to preprocessing and labeling in this case. A wide range of shape recognition approaches have been proposed, such as structural (e.g. methods organizing local features into graphs [5] trees [6] or strings [7] fuzzy or probablistic (e.g. relaxation methods [8]) statistical (e.g. methods based on moments [9] methods that work on some transform domain (e.g. Fourier [10] or Hough [11] and methods based on Neural Networks [12, 13] An important class of contour tracking and matching methods relies on physical models of the deformation and is based ....

Bir Bhanu and Olivier D. Faugeras. Shape Matching of Two-Dimensional Objects. IEEE Transactions on Pattern Analysis and Machine Intelligence, 6(2):137--156, 1984.

....the source and target images, match estimation must deal with potentially large, locally discontinuous offsets, with areas of occlusion and disocclusion and with differences in lighting and surface reflectance. Scalespace methods[1,2] non linear local transform methods[3] feature based methods[4,5], and non linear distance measures[5] have all been used in motion stereo to address similar difficulties. Unfortunately, their underlying assumptions (about the stability of shape and the rigidity of the motion, or about the lack of rotation expansion compression) will be violated within the ....

B. Bhanu, O. Faugeras, "Shape Matching of Two-Dimensional Objects," IEEE Trans. PAMI, vol 6, pp. 137--156, 1984.

....From another view, image segmentation is a labeling problem; the ideal segmentation should be consistent or nearest to the one with maximum likelihood. This has led to various relaxation labeling or stochastic labeling methods that are related to general optimization algorithms. Bhanu and Faugeras [4] regarded the shape matching as segment matching problem, and maximized a criterion function based on the ambiguity and inconsistency of classification. In their approach, a contour model was used to do the matching and define the criterion function. Hummel and Zucker [25] pointed out that a ....

Bir Bhanu and Olivier D. Faugeras. Shape matching of twodimensional objects. PAMI, 6(2):137--156, February 1984.

....1983) who first showed that the probabilistic relaxation scheme of (Rosenfeld et al. 1976) could be interpreted as the minimisation of an energy function based on label probabilities and a set of heuristically defined support functions. Following on from this, Faugeras and Berthod, 1981) and (Bhanu and Faugeras, 1984) developed similar energy based criteria which maximised consistency while minimising ambiguity in the labelling. The later work of (Kittler and Hancock, 1989) demonstrated how the process of probabilistic relaxation could be viewed entirely in terms of probability distributions, specifying ....

....may be interpreted as the probability that node u is matched to node v. If Equation 1. 3 is to satisfied we can no longer simply update the matching vector ae by the gradient of the matching criterion since we have no guarantee that this will lead to an allowed value of the matching vector (see (Bhanu and Faugeras, 1984; Zucker and Mohammed, 1978; Faugeras and Berthod, 1981; Luo et al. 1988) Instead we must use a gradient projection method. In this method the constraints in Eqn. 1.3 are viewed as defining a sub space in the labelling space which contains only the allowed values of ae. We then project the ....

[Article contains additional citation context not shown here]

Bhanu, B. and Faugeras, O. D. (1984). Shape matching of two-dimensional objects. IEEE Transations on Pattern Analysis and Machine Intelligence, 6(2):137--155.

....search methods, with various modifications to reduce the computational complexity. More recently other methods, based on energy minimisation, have been proposed, using simulated annealing [1, 20] mean field theory [15] or deterministic annealing [5, 6, 18, 23, 29] and relaxation labelling [2, 3, 11, 12, 19, 25, 27]. The relaxation approach in particular has the advantage that it converts the NP complete problem into one of polynomial complexity. The derivation of the formulae used in the relaxation labelling methods has been until recently essentially heuristic. This lack of a solid theoretical underpinning ....

B. Bhanu and O.D. Faugeras. Shape matching of two-dimensional objects. IEEE Trans. Pattern Analysis and Machine Intelligence, 6:137--156, 1984.

....analysis, stereo correspondence, feature labeling, object recognition, scene understanding, etc. can be seen as examples of ARG matching. Inexact matching between images and models is commonly performed by explicit search approaches or by using implicit search methods such as relaxation labeling [27]. A goodness of fit is usually formulated to measure the optimality of matching. In the following, we will present the proposed approach firstly at the representational level, secondly at the computational level and thirdly at the algorithm level and analyze the complexity of the algorithm. After ....

B. Bhanu and O. D. Faugeras. "Shape matching of two-dimensional objects". IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-6(2):137--155, March 1984.

....one. This is the tactic employed by heuristic search termination (Grimson 1990) which terminates the search as soon as a solution meeting some minimum requirement has been found. Relaxation labeling has also been used as a way to shorten the search while accepting a sub optimal result (e.g. Bhanu and Faugeras 1984; Bray 1990; Kitchen 1980) 3.3.2 Transformation space search The generalized Hough transform is an example of a method that searches transformation space. An array of bins, indexed by parameters of object pose, is first initialized as empty. Then, for each possible match between one image feature ....

Bhanu, B. and Faugeras, O.D. 1984. "Shape matching of two-dimensional objects." IEEE Trans. Patt.

....feature labeling, object recognition, scene understanding, etc. can be seen as examples of ARS matching. Inexact matching between images and models is commonly performed by explicit search approaches [1, 6, 7, 10, 13, 14, 20, 23] or by using implicit search methods such as relaxation labeling [4]. A goodness of fit is usually formulated to measure the optimality of matching. This paper is organized as follows. Section 2 discusses proper selection of invariant properties and relations and introduces the ARS representation. Section 3 formulates the problem of optimal ARS matching, the ....

B. Bhanu and O. D. Faugeras. "Shape matching of two-dimensional objects". IEEE Transactions on Pattern Analysis and Machine Intelligence, 6(2):137--155, March 1984.

....to preserve their order. But in the absence of any ordering information (like in stereo matching of many small fragments of curves) the matching algorithm may be called spread primitive matching . In this category we find algorithms that seek isomorphism between attributed relational graphs [6, 27, 10], and algorithms that look for the largest set of mutually compatible matches. Here, compatibility means an agreement on the induced coordinate transformation, and a few techniques exist to find the largest set of mutually compatible matches (e.g. clustering in Hough space [37] geometrical ....

Bhanu B. and Faugeras O., "Shape matching of two dimensional objects", PAMI 6, 137-155, 1984.

....can be nonlinear in f , and often measures the violation of a certain constraint incurred by the solution f . This linear combination of (nonlinear) potential functions is not an unusual form and can be found in matching and recognition works such as [11] 13] 9] 16] 18] 34] 27] [4], 41] 12] 26] 38] 39] 21] Note that when E(f j ) takes the linear form, multiplying by a positive factor 0 does not change the minimal configuration arg min f E(f j ) arg min f E(f j ) 3) Because of this equivalent, an additional constraint should be imposed on for ....

Bhanu, B. and Faugeras, O. D. (1984). "Shape matching of two-dimensional objects". IEEE Transactions on Pattern Analysis and Machine Intelligence, 6(2):137-- 155.

No context found.

Bhanu B. and Faugeras O., \Shape matching of two dimensional objects", PAMI 6, 137-155, 1984.

No context found.

B. Bhanu and O. D. Daugeras. Shape Matching of Two-Dimensional Objects. IEEE Trans. on Pattern Analysis and Machine Intelligence, 6(2):137--156, 1984.

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