Pollard, S.B., Mayhew, J.E.W., and Frisby, J.P. 1985. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... different points, i.e. shiftable windows [2, 18] windows with adaptive sizes [84, 60, 124, 61] and windows based on connected components of constant disparity [22] Threedimensional support functions that have been proposed include limited disparity difference [50] limited disparity gradient [88], and Prazdny s coherence principle [89] Aggregation with a fixed support region can be performed using 2D or 3D convolution, C(x, y, d) w(x, y, d) C 0 (x, y, d) 2) or, in the case of rectangular windows, using efficient (moving average) box filters. Shiftable windows can also be ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....established based on the selection of the best matching signatures. Typically, only the features are matched producing sparse disparity maps. The Marr s cohesivity principle (based on the surface continuity assumption) have been improved in the PMF algorithm designed by Pollard, Mayhew and Frisby [50], where the cohesiveness is represented by the condition of disparity gradient limit. In early 90 s, the researches focused on a modification of Marr s theory, where disparity would be estimated in all pixels directly. In this moment, the research in stereo matching has been divided into two main ....

....cost is improved based on the contributions of points from a defined neighbourhood corresponding with high probability to the same scene patch: support region. A support region can be represented by various kinds of adaptive windows, limited disparity di#erence [24] limited disparity gradients [50]. Di#erent ways of aggregation are iterative di#usion [59] or belief propagation [63] The local methods are in general very fast, they work only on a small area of an image. They do not produce any artifacts as global methods, however if the selection method is not properly formulated they could ....

[Article contains additional citation context not shown here]

Stephen B. Pollard, John E. W. Mayhew, and John P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

.... at different points (shiftable windows) 2, 14] windows with adaptive sizes [53, 40, 78, 41] and windows based on connected components of constant disparity [17] Three dimensional support functions that have been proposed include limited disparity difference [33] limited disparity gradient [56], and Prazdny s coherence principle [57] Aggregation with a fixed support region can be performed using 2D or 3D convolution, or, in the case of rectangular windows, using efficient (moving average) boxfilters. Shiftable windows can also be implemented efficiently using a separable sliding ....

S. B. Pollard, J. E. W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449-470, 1985.

....gradient and local neighborhood relations. The contours are defined by the zero crossings of the second derivative in the direction of the local gradient of the smoothed intensity function. Constraints: epipolar geometry, 2] uniqueness, 1] disparity continuity, 1] dis parity gradient, [4]. M5: A method using disparity histograms as suggested by Shirai and Nishimoto, 5] Candidates are matched with regard to global and local disparity histograms. The distribution of possible disparity values in a local neighborhood of a matching candidate is computed as disparity histogram in ....

Pollard S.B., Mayhew J.E.W., Frisby J.P.: PMF: a stereo correspondence algorithm using a disparity gradient limit. Perception, 14,449-470.

....(annealing, graph optimization) are still computationally expensive. Neighborhood interactions are also very commonly found in biological literature and several cooperative schemes have been proposed, with di#erent facilitation inibhition strategies along the spatial and disparity coordinates [18, 21, 20]. For the sake of computational complexity we adopt a spatial only facilitation scheme whose principle is to reinforce the output of units at locations whose coherent neighbors (tuned for the same disparity) are active. This scheme can be implemented very e#ciently by convolving each of the ....

S. Pollard, J. Mayhew, and J. Frisby. Pmf: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....sense. The Class 1 and 2 ambiguities require a prior model to regularize the stereoscopic matching problem. Usually, some form of uniform continuous prior is imposed. The early solutions based on the heuristic of cohesiveness [8] culminated in the classical algorithm by Pollard, Mayhew and Frisby [9]. Although some interest remained [10] many authors departed from this path and tried to pose the problem in the framework of statistical estimation theory with a probabilistic continuity prior. The ML estimators of early 90 s, based on dynamic programming, were later put on a more sound basis by ....

Pollard, S.B., Mayhew, J.E.W., Frisby, J.P.: PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception 14 (1985) 449--470

....the uncertainty in depth. This leads to a conflict: short baselines simplify the matching process, but produce uncertain results; long baselines produce precise results, but complicate the matching process. One popular set of approaches for dealing with this problem are relaxation techniques [6, 9]. These methods are generally used on a pair of images; start with an educated guess for the correspondences; then update them by propagating constraints. These techniques don t always converge and don t always recover the correct correspondences. Another approach is to use multiple images. ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. Pmf: A stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449--470, 1985.

....stereo algorithm. The original stretch correlation algorithm [11] used area based correlation to guide the matching of the left and right edge features. The algorithm exploited epipolar geometry, to reduce the correspondence problem to a 1 D search, together with a disparity gradient limit [16]. The temporal stereo algorithm improved on the conventional stretch correlation algorithm by making the epipolar search bands adaptive; calculating search band sizes and locations based on disparity information obtained from the previous image pair. Temporally seeding using disparity values, ....

....as the disparity gradient can still be applied at all image scales: According to [1] a change in the viewing distance scales both dot separation and disparity and, under experimentation, the disparity gradient seemed to remain constant over a wide variety of scales. This view is also supported by [16] which also says that the region of disparity gradient support should be set to be large enough to provide good disambiguation while not being so large that processing time is spent needlessly. One of the other advantages of the CTF approach being used at the beginning of the sequence is that a ....

Pollard S.B. Mayhew J.E.W. and Frisby J.P. PMF: A Stereo Correspondence Algorithm using a Disparity Gradient Limit. Perception, 1985, Vol. 14, pp. 449-470.

.... aggregation has been done using square windows (traditional) Gaussian convolution, multiple windows anchored at different points [15] and windows with adaptive sizes [17] Three dimensional support functions that have been proposed include limited disparity difference, limited disparity gradient [25], and Prazdny s coherence principle [26] BMVC99 316 Sub pixel disparity estimates can be computed by fitting a curve to the matching costs at the discrete disparity levels [22, 17] This provides an easy way to increase the resolution of a stereo algorithm with little additional computation. ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....in the Z. other image Fua, 1993 , where the number of pairs Z. of features to be considered becomes high. 2 Feature based methods use sets of pixels with similar attributes, normally either pixels belonging to edges Z Kim and Aggarwal, 1987; Marr and Poggio, 1979; Mousavi and Schalkoff, 1994; Pollard et al. 1981 Z or the corresponding edges themselves Ayache and Faverjon, 1987; Cruz et al. 1995a,b; Hoff and Ahuja, 1989; Medioni and Nevatia, 1985; Pajares, 1995 . As Z. shown in Ozanian, 1995 , these last methods lead to a sparse depth map only, leaving the rest of the surface to be reconstructed by ....

Pollard, S.B., Mayhew, J.E.W., Frisby, J.P., 1981. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception 14, 449--470.

....stereo correspondence calculations. 43 6.4 Stereo matching and production of a random dot stereogram Given a pair of parallel camera images, such as provided by the rectification discussed in this report, much more sophisticated stereo matching can be performed. As discussed in [13] and [14], this can at first be a constrained inter scanline search by cross correlation. Vertical edges crossing the scanline can be identified from zero crossings in the gradient d I dx 2 2 and candidate matches identified as before by cross correlation. Rather than simply selecting the match with the ....

S Pollard, J Mayhew, J Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449-470, 1985

....[5] Several points on one epipolar line might have the same intensity structure and a unique solution to the stereo matching problem cannot be given. This implies that additional information is needed. Typically, constraints such as figural continuity, smoothness, and maximum disparity gradient [6] are imposed. Assumptions of the statistics of the disparity gradient and Maximum A Posteriori (MAP) estimation has been used to constrain the solutions. This leads to the regularization scheme, as it was earlier proposed by Grimson [7] If the possibility of discontinuities is taken into account ....

S. B. Pollard, J. E. W. Mayhew, J. P. Frisby, "PMF: A stereo correspondence algorithm using a disparity gradient limit", Perception, Vol. 14, pp. 449-470, 1985.

....B 1 lie close enough on the same epipolar line in a certain order (e.g. A 1 is closer to the epipole E 1 than B 1 ) in the first retina, then the images of the corresponding 3 D points in the second retina lie in the same order. This condition can also be replaced by the disparity gradient limit [PMF85] which may be closer to the human visual 4 system limit. Since the disparity gradient is related to local surface orientation, this is equivalent to setting a maximum angle between the local orientation of the surface and the optic ray. A better method to compute the derivatives of disparity ....

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF : a stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449--470, 1985.

....through an appropriate regularization term. Over the years, numerous algorithms for stereo vision have been proposed, which use dioeerent strategies. # Feature based: Those algorithms establish correspondences between some selected features extracted from the images [22] such as edge pixels [41, 43], line segments [5, 31, 32] or curves [11, 39, 47] Their main advantage is to yield accurate information and to manipulate reasonably small amounts of data, thus gaining in time and space complexity. Their main drawback is the sparseness of the recovered depth information. This class of methods ....

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF : a stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449470, 1985.

....through an appropriate regularization term. Over the years, numerous algorithms for stereo vision have been proposed, which use di erent strategies. Feature based: These algorithms establish correspondences between some selected features extracted from the images [15, 22] such as edge pixels [37, 38], line segments [6, 29, 30] or curves [10, 33, 43] Their main advantage is to yield accurate information and to manipulate reasonably small amounts of data, thus gaining in time and space complexity. Their main drawback is the sparseness of the recovered depth information. This class of methods ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby, PMF : a stereo correspondence algorithm using a disparity gradient constraint, Perception 14, 1985, 449-470.

....f p r g, i.e. to resolve the ambiguity, we rely on the available contextual information, as shown in the following. In the literature, ambiguity is typically resolved by imposing constraints on the disparity field, such as uniqueness, ordering (or monotonicity [7] and smoothness [20] 8] [23], 3] 22] 21] Note that also other algorithms that make use of vectorial local descriptions ( 14] 11] need to impose constraints on the disparity field: the notion of closeness in the representation space is not itself sufficient for a reliable match. The main novelty of our approach is ....

S.B. Pollard, J.E. Mayhew, and G.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

.... surfaces) or three dimensional in (x, y, d) space (allowing slanted surfaces) Two dimensional evidence aggregation has been done using both fixed square windows (traditional) and windows with adaptive sizes [Arn83, KO94] Three dimensional support functions include limited disparity gradient [PMF85] Prazdny s coherence principle [Pra85] which can be implemented using two di#usion processes [SH85] local winner take all [YYL93] and iterative (non linear) evidence aggregation [SS96] In this paper, our initial evidence aggregation uses an iterative technique, with estimates being refined ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....and dislocation in edge detection can be appropriately coped with by adding a constant term Deltad to Deltadisp 0 (see (2) which is usually in the order of 1 or 2 pixels. 1 For details of the implementation, see [8] 3.2. 2 Reinforcing the shape similarity constraint Most algorithms such as [12][13] which make use of DG limit weight the support (or vote) by the inverse of distance. Their argument was that when Deltadisp 0 ( Deltadist) increases, false matches may also get maximal global support by chance. Thus the possibility of choosing false match matches increases, especially when ....

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....displacement vector along the contours, and thus the deformation from one contour to the other. It is similar to the one used by Hildreth [5] to compute the velocity eld between images, and can be related to the gural continuity and disparity gradient constraints introduced by Mayhew and Frisby [8]. Finding the optimal deformation measure can now be stated as a variational problem where the following functional has to be minimized: D(fl) Z t 1 t 0 F (fl; t)dt where : F (fl; t) k dD(fl(t) dt k 2 A classical approach involves the resolution of the Euler Lagrange equation: F fl ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF - a stereo correspondence algorithm using a disparity gradient limit. Perception, 14(4):449 470, 1985.

....from different viewpoints. This is all in contrast to methods based on optical flow [2] which of necessity utilize surface smoothness and temporal coherence constraints to solve the aperture problem. In addition there is a long tradition of incorporating surface constraints in stereo matching [39, 22, 35]. In the latter cases it is clearly difficult to generalize these already somewhat unwieldy methods to complex scenes and multiple images with general camera motion. The benefit of independently computing features in 3D is that the algorithms are relatively clean and scale well to complex data ....

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....that are to be put into correspondence to make educated quesses as to what their match should be. Then, let the matches reorganize themselves by propagating some of the constraints described in section 3.1.2. We refer to the Marr Poggio algorithm [MP79] and Pollard, Mayhew, Frisby algorithm [PMF85] as example of application of those ideas to stereo vision. Dynamic programming The matching of an image pair can be done by matching proper epipolar line pairs separately. The problem of matching two intensity lines can be considered as problem of mapping two time variant functions by rescaling ....

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....stable matchings including the X dominant matching, the FX dominant matching, and the maximum stable monotonic matching. It was shown that all of them are subsets of stable complete matching. It is interesting to note that the idea of stability is present intrinsically in the famous PMF algorithm [13]. We saw that stable complete matching can be found in O(n 2 ) time where n is the image width in pixels. Stable monotonic matching (SMM) imposes ordering constraint on the solution. We have presented a simple sub optimal O(n 2 ) algorithm for SMM (Algorithm 2) which works much like dynamic ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....area spans a 2D area (d=const. in the r c d space. This 2D local support area corresponds to the fronto parallel plane assumption. However, sloping and more general surfaces require using a 3D area in the disparity space for local support. Many 3D local support assumptions have been proposed [9] [25], 26] 12] Kanade and Okutomi [12] present a detailed analysis of the relationship and differences among them. For simplicity we use a box shaped 3D local support area with a fixed width, height and depth, but a different local support area could be used as well. Current Element, r, c, d) ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby, "Pmf: A stereo correspondence algorithm using a disparity gradient limit," Perception, vol. 14, pp. 449-470, 1985.

....given pixel, assuming that the grey levels are similar. They yield dense depth maps, but fail within occluded areas and or poorly textured regions. Several correlation related measures have been proposed, the Sum of Squared Di erences (SSD) measure being a choice adopted most widely. Feature based [10, 16, 15, 14] algorithms match local cues (e.g. edges, segments, corners) and can provide robust, but sparse, disparity maps requiring interpolation. These algorithms depend on feature extraction to locate reliable features in the two images. This paper presents a new symmetric, multi window algorithm ....

S. B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449-470, 1985.

....image and then correspondences for each of them are sought for in the second image, generally via multi scale techniques. In the second approach, which the present work addresses, features are detected independently in both images and then matched up usually by relaxation (see, e.g. the classic [6]) Incidentally, recent state of the art work on the fundamental matrix estimation [14, 11] follows this latter avenue for achieving initial correspondences. This paper proposes an new neat and simple algorithm for achieving feature correspondence across pairs of images. Despite the well known ....

....C(i,J(i) 1; end end end Figure 5: MATLAB code of the SVD stereo correspondence method proposed in this paper. possible to embed any non pairwise constraint, such as the disparity gradient or the geometric ordering constraint[6] . Some of these problems could be bypassed, though. If, for instance, in a two stage approach a few definitely good matches are known beforehand, one could recover a very coarse epipolar geometry and embed the deviation from epipolar lines in the correspondence strength function in Equation (2) ....

S. Pollard, J. Mayhew, and J. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, (14), 1985.

....to impose a constraint of smoothness on the resulting disparity solution. While the latter constraint breaks down at the surface boundaries with depth discontinuities, window correlations give erroneous signals at regions of occlusion and depth discontinuities. The edge based stereo methods [4, 5, 14] which try to match compatible feature points in the left and the right images using smoothness constraint along gurally continuous sections are more reliable; but such methods can only provide sparse disparity maps. On the other hand, methods of computation of optic ow also require a smoothness ....

S.B. Pollard, E.W. Mayhew, and Frisby J.P. PMF: A Stereo Correspondence Algorithm Using a Disparity Gradient Limit. Perception, 14:449-470, 1985.

.... surfaces) Two dimensional evidence aggregation has been done using square windows (traditional) Gaussian convolution [14] and windows with adaptive sizes [10] Three dimensional support functions that have been proposed include limited disparity difference [8] limited disparity gradient [12], and Prazdny s coherence principle [13] which can be implemented using two diffusion processes [17] Some matching costs, such as correlation and nonparametric measures are defined over a certain area of support, and thus combine the cost and aggregation steps into one. Measures that can be ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....try to find matches over a wide range of disparities (on the order of hundreds of pixels) face difficulties in trying to guarantee a unique match based on local attributes of features, like contrast and orientation. One solution to this problem is to use attributes of nearby features [3] 46] [40] and another is to alter the control strategy. Since we are interested in finding roughly contiguous 3D regions to select out groups of image features that are likely to come from a single object, we use a control method fixates a target, searches for matching features in some narrow disparity ....

.... similar to the working of the human stereo system where the fusible range of disparities is restricted around the fixation point (Panum s area) The stereo algorithm implemented here is a modified version of Grimson s stereo matcher ( 20] It is similar to earlier stereo algorithms [3] and [46] [40], 37] 41] and uses ideas about the human stereo system, Panum s area and the role of eye movements in stereopsis as discussed in [37] 41] 24] and [50] The stereo algorithm does the following: ffl decides what features to match in the two images, ffl decides how the matching is to be done, ....

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby, "PMF: A stereo correspondence algorithm using a disparity gradient limit", Perception 14, 449--470, 1985.

....of the tokens in each image [32] Burt and Julesz [32] provide evidence supporting the claim that, the human visual system applies a disparity gradient limit to the oriented edge points detected on the retina, and the measured human visual system value for the DGL is close to unity. Pollard et al. [56] suggest that for most natural scene surfaces, including jagged ones, the disparity gradient between correct matches is usually 1, whereas it is very rare among incorrect matches obtained for the same set of images. Dhond and Aggarwal in their review paper [3] draw a picture for the original de ....

....natural scene surfaces, including jagged ones, the disparity gradient between correct matches is usually 1, whereas it is very rare among incorrect matches obtained for the same set of images. Dhond and Aggarwal in their review paper [3] draw a picture for the original de nition and calculation [32, 56] of disparity gradient limit . Under the assumption of zero vergence [36] Jones[4] related the angular polar coordinates L and R of two lines in each image by 4(tan L tan R ) 2 L 2 dg ( tan L tan R ) 2 4) The above mentioned constraints (similarity, epipolar constraint ....

[Article contains additional citation context not shown here]

S. B. Pollard, J. Mayhew, and J. Frisby, \Pmf: A stereo correspondence algorithm using a disparity gradient constraint," Perception, vol. 14, pp. 449-470, 1985.

....image [14, 15] A similar vector is estimated on the other image, in order to compute matching. Matching image features is generally more robust; the related class of algorithms is called feature based. In the present context, the term features indicates physically meaningful cues, such as edges [16, 17, 18, 19, 20]; segments (collinear connected edges) 21] and corners (where two edges cross) 22] Features can be extracted by bandpass filters, derivative operators or ad hoc non2 linear operators. The local phase of the image signal computed via Fourier or Gabor transforms has been matched also [23, ....

.... that yields the maximum similarity score between the left and the right elements [8, 4, 3, 9, 10, 11, 13] With relaxation based methods the elements are joined by weighted links; initial weights are iteratively updated by propagating constraints, until some equilibrium configuration is reached [1, 16, 17, 18, 22]. Dynamic programming techniques adopt a cost function, which embeds the constraints and is minimised to get the best set of matches [20, 19, 10, 12, 26, 2] The solution is a path in the match space [2, 12] or the disparity space [10] Usually, the energy functional is derived using Bayesian ....

S. B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449-- 470, 1985.

....a full model feature be observed. More typically, due to occlusion and noise, only a portion of the feature will be observed; e.g. a small surface patch or an edge fragment. In our work we consider either type of input data: either surface data from our in house (Oxford NEL) laser range finder [31,33], or 3D edge fragments from a stereo vision system [29] Surface data are represented by a 3D position and unit surface normal, s = p; n) This representation is rather conservative since the dense range maps supplied by the Oxford NEL sensor give information about surface extent and ....

....and the network evaluation throughout a MIMD processor network [7] A Binary geometry constraints This appendix specifies the feature constraints (i.e. the f k ) used in our system. Two types of data may be input: either surface data from our in house (Oxford NEL) laser range finding system [31,33], or three dimensional edge fragments from a stereo system [29] Below we give sets of feature constraints for each of these data types. Demonstrations of the system working with both types of data are given in Sections 4 to 5. Surface data are represented by a 3D position and a unit surface ....

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

....a large variety of image tokens such as edges, points of interest, line segments, contours, regions and even colour. This work does not concern itself primarily with describing the large body of work employing edge points or points of interest which makes up the bulk of early stereoscopic work [6, 8, 14, 30, 40, 49, 50, 53, 63, 64, 65, 67, 78, 79]. However elements of all of these references will be used to illustrate generic aspects of correspondence processes. Low level features such as edges exhibit two main disadvantages as matching tokens. First, the hundreds of edge points extracted demand very large computational resources which ....

.... pair of matches, the disparity gradient is defined as the ratio of the difference in disparity of the two matches to the average separation of the tokens in each image [14] Originally, a limit on the disparity gradient between neighbouring matches was employed as a binary continuity constraint [65, 67]. However [49, 65] reformulated the DGL as a unary geometric constraint on orientated features. Under the assumption of zero vergence, the angular polar coordinates l and r of two lines in each image are related by the following 4(tan l 0 tan r ) 2 L 2 dg i (tan l tan r ) 2 ....

[Article contains additional citation context not shown here]

S.A. Pollard, J.E.W. Mayhew, and J.P. Frisby. "PMF: A Stereo Correspondence Algorithm using a Disparity Gradient Limit". Perception, 14:449--470, 1985.

....faithful representation of a contour which would otherwise be represented by a polygonal chain. In this paper we are not considering closely spaced views where snake tracking is viable. Basic criteria for curve matching in stereo pairs were established by Pollard et al. in the PMF Stereo Algorithm [16]. Their primary criteria were the epipolar and ordering constraints, but these were supplemented by figural continuity to overcome problems where curves coincide with epipolar lines (since if the curve coincides with an epipolar line, point correspondences cannot be determined) Zhang and ....

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449--470, 1985.

.... Many computational models for biological motion processing (Reichardt, 1961; Hildreth, 1984; Watson and Ahumada, 1985; van Santen and Sperling, 1985; Adelson and Bergen, 1985; Heeger, 1987; Grzywacz and Yuille, 1990) and stereo vision (Marr and Poggio, 1976; Marr and Poggio, 1979; Prazdny, 1985; Pollard, Mayhew and Frisby, 1985; Sanger, 1988; Qian and Sejnowski, 1989; Yeshurun and Schwartz, 1989) have been proposed, but few dealt with these two visual functions at the same time. Although it is clear that at an abstract level both motion and stereo vision can be formulated as solving a correspondence problem (see Marr ....

....energy models 1 (Adelson and Bergen, 1985; Watson and Ahumada, 1985; van Santen and Sperling, 1985; Emerson et al. 1992) Until recently, most models of stereopsis, on the other hand, cannot be said to be truly biological. Some (Marr and Poggio, 1976; Marr and Poggio, 1979; Prazdny, 1985; Pollard et al. 1985; Qian and Sejnowski, 1989) require very sharply disparitytuned units and use explicit matching of fine image features (see the Discussion) Others (Sanger, 1988; Yeshurun and Schwartz, 1989) contain certain mathematical operations (such as the explicit extraction of complex phases of stimuli) ....

Pollard, S. B., Mayhew, J. E. and Frisby, J. P. (1985). PMF: A stereo correspondence algorithm using a disparity gradient limit, Perception 14: 449--470.

....and the constraint look up tables will be different. An accurate geometric representation of the model is stored in memory and recovered scene data is compared systematically to this. In this work, the matching is performed using the edge information recovered by a structure from stereo algorithm [Poll85]. The basic recognition procedure is outlined: 1) Build an accurate geometric representation of the model in memory. 2) Recover the data edges of an object in the scene of interest using a reliable 3D reconstruction algorithm. 3) Place the data in the form of an interpretation tree (IT) and from ....

Pollard, S.B., Mayhew, J.E.W., Frisby, J.P., "PMF: A Stereo Correspondence Algorithm Using a Disparity Gradient Limit", Perception, vol. 14, pp. 449-470, 1985.

No context found.

S B Pollard, J E W Mayhew, and J P Frisby. Pmf: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

No context found.

. PMF:A stereo correspondence algorithm using a disparity gradient limit. Perception, 14, 449--470.

....occur due to noise in images or the lack of robustness of edge operator. The occurances in this case are always at the end points of an edge. Therefore, the monotonicity constraint is suspended for those points whose correspondences are end points of an edge. Disparity Gradient In the PMF stereo [14], disparity gradient has been exploited to give neibourhood support to a match. For two matches M (j 0 ; m 0 ; n 0 ) 1 and M (j 1 ; m 1 ; n 1 ) 1, their disparity gradient ffi d is defined as ffi d = 2k(p r m 0 n 0 Gammap l i 1 j 1 ) Gamma(p r m 0 n 0 Gammap l i 0 j 0 )k k(p ....

S. B. Pollard, J. E. W. Mayhew and J. P. Frisby, PMF: A stereo correspondence algorithm using a disparity gradient limit, Perception, 1985, 14:449-470

No context found.

Pollard, S.B., Mayhew, J.E.W., and Frisby, J.P. 1985. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470.

No context found.

Pollard, S.B., Mayhew, J.E.W., Frisby, J.P.: PMF: a stereo correspondence algorithm using a disparity gradient limit. Perception 14 (1985) 449--470

No context found.

S.B. Pollard, J.E. Mayhew, and G.P.Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449#470, 1985.

No context found.

S. Pollard, J. Mayhew, and J. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, (14), 1985.

No context found.

S. B. Pollard, J. E. W. Mayhew and J. P. Frisby, "PMF: A stereo correspondence algorithm using a disparity gradient limit", Perception, vol.14, pp.449-470, 1985.

No context found.

S. B. Pollard, J. E. W. Mayhew, and J. P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

No context found.

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449-470, 1985.

No context found.

S.B. Pollard, J.E.W. Mayhew, and J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient constraint. Perception, 14:449--470, 1985.

No context found.

S. B. Pollard, J. E. W. Mayhew and J. P. Frisby, "PMF: a stereo correspondence algorithm using a disparity gradient constraint," Perception 14 (1985) 449--470.

No context found.

, 1980. #14# S.B. Pollard, J.E.W. Mayhew, and J.P.Frisby. PMF : a stereo correspondence algorithm using a disparity gradient constraint. Perception,

No context found.

S.B. Pollard, J.E.W. Mayhew & J.P. Frisby. PMF: A stereo correspondence algorithm using a disparity gradient limit. Perception, 14:449--470, 1985.

No context found.

S. Pollard and J. Mayhew and J. Frisby. "PMF: a stereo correspondence algorithm using the disparity gradient limit." Perception, 14: 449-470, 1985.

First 50 documents  Next 50

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