M. Okutomi and T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

.... or three dimensional in x y d space (supporting slanted surfaces) Two dimensional evidence aggregation has been implemented using square windows or Gaussian convolution (traditional) multiple windows anchored at 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 ....

....[41] wavelet phase none phase matching Birchfield and Tomasi [12] shifted abs. diff none DP Marr and Poggio [73] binary images iterative aggregation WTA Prazdny [89] binary images 3D aggregation WTA Szeliski and Hinton [114] binary images iterative 3D aggregation WTA Okutomi and Kanade [84] squared difference adaptive window WTA Yang et al. 127] cross correlation non linear filtering hier. WTA Shah [103] squared difference non linear diffusion regularization Boykov et al. 22] thresh. abs. diff. connected component WTA Scharstein and Szeliski [97] robust sq. diff. iterative 3D ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. IJCV, 7(2):143--162, 1992.

....refers to the manner in which the error function over the search space is computed or accumulated. The most direct way is to apply search windows of a fixed size over a prescribed disparity space for multiple cameras [OK93] or for verged camera configuration [KWZK95] Others use adaptive windows [OK92], shiftable windows [Arn83, BI99, TSK01] or multiple masks [NMSO96] Another set of methods accumulates votes in 3D space, e.g. the space sweep approach [Col96] and voxel coloring and its variants [SD97, SG99, KS99] More sophisticated methods take into account occlusion in the formulation, for ....

....Shan [ZS00] which starts with point matches and grows matching regions around these points. In our approach, however, there is no requirement to grow existing regions; instead, the most confident pixels are simply selected at each iteration. Our idea of variable window sizes is also similar to [OK92]. However, we adopt a highest confidence first approach [CB90] to choosing a window size rather than testing at each pixel location all the windows sizes in order to select an optimal size. Results of using the incremental selection approach can be seen in Figure 10, Figure 12 (for sequence shown ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, April 1992.

....two dimensional at a fixed disparity, or three dimensional in xy d space. Two dimensional evidence aggregation has been implemented using square windows or Gaussian convolution (traditional) multiple windows anchored 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 ....

M. Okutomi and T Kanade. A locally adaptive window for signal matching. IJCV, 7(2): 143-162, 1992.

....in the other image. The aggregation method refers to the manner in which the error function over the search space is computed or accumulated. The most direct way is to apply search windows of a fixed size over a prescribed disparity space for multiple cameras [19] Others use adaptive windows [18], shiftable windows [6, 27] or multiple masks [17] Another set of methods accumulates votes in 3D space, e.g. the space sweep approach [9] and voxel coloring and its variants [22, 25, 14] Once the initial or aggregated matching costs have been computed, a decision must be made as to the ....

....and Shan [28] which starts with point matches and grows matching regions around these points. In our approach, however, there is no requirement to grow existing regions; instead, the most confident pixels are simply selected at each iteration. Our idea of variable window sizes is also similar to [18]. However, we adopt a highest confidence first approach [8] to choosing a window size rather than testing at each pixel location all the windows sizes in order to select an optimal size. Results of using the incremental selection approach can be seen in Figure 9. While it generally interpolates ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. Intl. J. Comp. Vision, 7(2):143--162, 1992.

....kinds of approaches to attenuate the influence caused by the phenomenon that the surface orientations locally deform the intensity patterns surrounding corresponding points. One approach is to adaptively change the window size according to the amount of depth variation within the window. Okutomi[12] proposed the adaptive window algorithm which selects an appropriate size of window by evaluating the local variation of both intensity and disparity. The algorithm can also deal with the depth discontinuities. However, since it requires the initial disparity estimates at numerous pixels within ....

M.Okutomi and T.Kanade. A locally adaptive window for signal matching. IJCV, 7:2:143-- 162, 1992.

....the previous movements of the camera and the computational delays. The approach has been tested with simulations and experiments, and several experimental results have been presented. Issues for future research include the automatic selection of the window size (an issue discussed in [11]) in order to select a window that has some texture variations, the computational improvement of the approach, and the implicit incorporation of the robot dynamics. 6 Acknowledgments This work has been supported by the Department of Energy (Sandia National Laboratories) through Contract ....

M. Okutomi and T. Kanade, "A locally adaptive window for signal matching," International Journal of Computer Vision, 7(2):143-162, 1992.

....for every pixel in an area, region based matching is the obvious choice. But to make sure the flow vectors are realistic, it is necessary to have enough texture in the region to ensure a good match. This can be achieved by using a dynamic region size similar to the approach of Okutomi and Kanade[12]. The idea is to use edge gradients as a measure of information and, at every pixel position, grow the support region until there is enough edge gradient information to justify matching. Furthermore, flow needs to be computed only for pixels contained within the moving object. Consequently, this ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, 1994.

....and textured images, this method gives good results. But, we observe that for small windows, it does not cover enough intensity variation, and for large ones, problems appear near disparity discontinuities and occluded regions. To alleviate these problems, T. Kanade and M. Okutomi propose in [10] to correct an initial disparity by adapting the window size to the local variations of the image intensity and disparity. In their article [1] R.Chung and R.Nevatia exploit structural features to recover discontinuity information and show the importance of image contours. Many other articles ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. Int. Journal of Computer Vision, 7(2):143--162, 1992.

.... Gamma (A) 25) in which (A) is the largest and Gamma (A) the smallest singular value of A (i.e. eigenvalue of A T A) Apart from a small condition number we prefer a large signal to noise ratio. To this end we require the smallest singular value to be larger than a fiducial threshold [67]: Gamma (A) 26) Considering this we have chosen an approach which takes all singular values into account by selecting parameters that minimise the so called Frobenius norm of A Gamma1 , defined as the sum of squares of all singular values: kA Gamma1 k 2 F = X ff 2 ff (A ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, 1992.

....correct match and then refine our match to a fraction of a pixel using an adaptive window algorithm. As a first pass we use a multi resolution image pyramid with a Moravec interest operator to get close to the right answer. We then apply an adaptive window algorithm due to Okutomi and Kanade. [20] [10] We present a framework for selecting subimages which unifies the one and two dimensional adaptive window algorithms they present, and test the algorithm on real images and artificial data sets. We report our observations about the workings of the adaptive algorithm. 1 1 Introduction ....

....of surrounding context is needed. The accuracy of the computed depth will then depend on how similar the surrounding context is to the image area we are trying to match. Kanade and Okutomi have proposed a method for computing optimal window size during the registration phase of stereo matching. [20] [10] That is, once an approximate match has been made, it can be refined, or registered, by looking at local image properties. They present a method for selecting the optimal rectangular windows for doing this on 2D images, and a method for selecting optimal windows for 1D signals which are ....

[Article contains additional citation context not shown here]

K. Okutomi and M. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7:143--162, 1992.

....it can result in a greater number of false positives in occlusion zones and increased smoothing of disparity across discontinuities, although the number of false negatives due to noise and outliers may decrease. Different approaches have been developed to tackle individual issues (for example, [9], 10] within the framework of linear correlation measures. For instance, Quam [10] addresses the 0162 8828 98 10.00 1998 IEEE ################ .# D.N. Bhat is with LG Electronics Research Center of America, 40 Washington Road, Princeton, NJ 08550. E mail: dbhat lgerca.com. # S.K. Nayar is ....

# M. Okutomi and T. Kanade, "A Locally Adaptive Window for Signal Matching," Int'l J. Computer Vision, vol. 7, no. 2, pp. 1,4991, 512, 1992.

....regions. To obtain sharp boundaries, single color pixels are matched. The cameras are decomposed to eight subsets and the disparity is retained from that one giving the best match. Okutomi and Kanade proposed a method that adaptively adjusts the matching window size based on image contents [15]. Various phenomena related to occluding contour are also studied in [3] and [14] Once it is known where the boundary is, matching can be applied to both its sides and the result with higher confidence retained. The difficulty is how to detect the boundary reliably. Even if it is done so, a ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. Int. J. Computer Vision, 7(2):143--162, 1992.

....of the sum of squared distances around a pixel. While, using windows as in SSD can increase the chance of finding the correct match, it also creates problems such as jagged edges in the depth map[4,15,16] Further attempts at local correspondence have also been attempted using adaptive window sizes[10,18], matching in the Fourier domain[9,12,25] and coarse to fine methods[3,6,13] One of the most promising methods has been using multiple cameras, or multi baseline stereo[14,15,19,21 23] Using multiple cameras can reduce the number of false matches without increasing the complexity of the method ....

Okutomi, M. and Kanade, T., "A Locally Adaptive Window for Signal Matching" International Journal of Computer Vision, 7:2, 1992:p. 143-162.

....all, it seems that the selected observation windows for both optimality criteria actually cover representative regions around all considered landmarks. 3. 6 Related Work Related work on the selection of suitable windows has been done in the context of matching stereo images by Okutomi and Kanade [9] and for junction localiza tion by Lindeberg [7] Lindeberg presented an approach for junction localization with automatic scale selection. This approach differs in various aspects from our approach. First of all, he selects so called scale space maxima of the operator by Kitchen and Rosenfeld ....

M. Okutomi and T. Kanade. A Locally Adaptive Window for Signal Matching. International Journal of Computer Vision, 7(2):143--162, 1992.

....laser and ultrasonic range finders, infra red obstacle detectors, GPS, Forward Looking Infra Red (FLIR) microwave, etc. provide other modalities with particular strengths and weaknesses. Issues for future research include the automatic selection of the feature window size (an issue discussed in [9]) in order to select a window that has some texture variations, simple model based techniques for distinguishing people or vehicles from background clutter, and the incorporation of the visual sensing data into larger, multisensor application systems for various transportation applications. ....

M. Okutomi and T. Kanade, "A locally adaptive window for signal matching," International Journal of Computer Vision, 7(2):143-162, 1992.

....two dimensional translational motion between two images from a sequence may be estimated by selecting a displacement which maximises the correlation value between corresponding image blocks. An alternative method to find the match position is to minimise the Sum of Squares Difference (SSD) 1][2]. Under certain assumptions both approaches yield similar results. Taken together, these algorithms are of basic importance to the field of image matching. The range of applications of image matching includes stereo vision (to compute depth) detection and tracking of moving objects (in, for ....

Masatoshi Okutomi and Takeo Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, February 1992.

.... in the image can be determined using various approaches: correlation [2] gradient [35, 42] spatio temporal filtering [26] or regularization [35, 54] For large image motion, multiresolution approaches are used to prevent local minima in the matching process [7, 62, 68] Adaptive window sizes [51] and quadtree splines [62] are used to treat different parts of the image with varying resolution. Affine flow or quadratic flow assumptions can be used to represent optical flow parametrically [7] Recovering the camera relationships for 2 frames can be solved using methods such as the ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, 1992.

....we now summarize the issues we left open in this report. Two parameters need to be specified for detection and tracking: the size of the window and the detection threshold. We have argued that windows should be as small as possible, compatibly with good noise rejection. However, it has been shown [Okutomi and Kanade, 1990] that a careful choice of the window size can improve performance considerably. It would be interesting to develop an inexpensive and automatic window size selection algorithm. The feature detection threshold was chosen in this report based on a histogram of the minor eigenvalues for the entire ....

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. In Proceedings of the Third International Conference on Computer Vision, pages 190--199, Osaka, Japan, December 1990.

No context found.

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, 1992.

No context found.

M. Okutomi and T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):1499--1512, 1992.

No context found.

M. Okutomi and T. Kanade, "A locally adaptive window for signal matching," International Journal of Computer Vision, 7(2):143-162, 1992.

No context found.

M. Okotumi, T. Kanade, "A locally Adaptive Window for Signal Matching", Int. Journal of Comp. Vision, Vol 7, no 2, pp 143-162, January 1992.

No context found.

M. Okutomi et T. Kanade. A locally adaptive window for signal matching. International Journal of Computer Vision, 7(2):143--162, 1992.

No context found.

M. Okutomi and T. Kanade. A Locally Adaptive Window for Signal Matching. International Journal of Computer Vision, 7#2#:143#162, 1992.

No context found.

M. Okutomi, T. Kanade, "A Locally Adaptive Window for Signal Matching", in Proc ICCV, Dec 1990.

First 50 documents

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