Y.G. Leclerc, "Capturing the local structure of image discontinuities in two dimensions, " in Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, CA, June, 1985, pp. 34--38.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....slowly, except at a limited set of boundaries. The user specifies a minimum spacing between adjacent boundaries. The edge finder treats regions smaller than this as texture. The parameters used in this paper impose a minimum boundary spacing of about 12 pixels. Previous texture edge finders (e.g. [14, 15, 24]) use similar constraints. I will make only weak assumptions about the fine scale variation: that its values are roughly clumped about zero and that significant correlations exist only between locations which are near one another (relative to the boundary spacing) 2 Robust estimation Robust ....

....of its neighborhood lies outside the corner. Rounding in the location estimates makes the corner points appear to be outliers; when there is a sharp change in scale, rounding the scale estimates has the same effect. Previous robust and texture estimators all exhibit this artifact, except for one [15] which uses many 2D operator shapes and seems very slow. To correct rounding, the new estimator first marks problem locations. A pixel s estimates are marked as wrong if it is near a significant change in scale (scale estimates for an opposite pair of neighbors differ by a factor of two) or its ....

Leclerc, Y. 1985. Capturing the Local Structure of Image Discontinuities in Two Dimensions. IEEE CVPR 34--38.

....sides of the line is h and H, respectively. Then this disc template is fitted to the data by adjusting (k, b, h, H) After convergence, significant difference of h and H indicates the existence of a step edge, whose position and orientation are determined by the line y=kx b. Leclerc and Zucker [15,16] fitted polynomial curves to the left and right hand profiles of a point on the surface, then the values and derivatives of the left and righthand profiles at this point are compared to decide if this point is a continuous point or not. Blake and Zisserman [17] and Terzopoulos [18] were the ....

....using TP NURBS had to tolerate the pole degeneracy, or use two or more pieces of surfaces and then glue them together. a) scattered data (d) fine tuning and fairing by adjusting weights Figure 6. spherical data (b) triangulation of the unit sphere domain (c) C 1 surface with preserved edges 16 Figure 7 (a) is the shaded display of the range image of an airplane, and (b) shows 900 scattered samples. We apply a Delaunay triangulation algorithm [28] to the scattered data and produce the initial triangular mesh (c) in which some triangle edges severely misalign at the boundary between ....

Y. Leclerc, Capturing the Local Structure of Image Discontinuities in Two Dimensions, IEEE CVPR, Jun. 1985

.... Although homogeneous regions are a useful description for a certain class of images, there is a much wider class that is more usefully described as having piecewise smooth image attributes, that is, attributes that are almost everywhere continuous and differentiable up to some specified low order [1, 4, 10, 12, 14, 16, 17, 19, 20, 26, 27]. For example, images of objects with piecewise smooth surfaces and albedos are well described in this manner. For simplicity, we shall restrict our discussion henceforth to a single image attribute, namely, intensity, although the partitioning technique developed here is directly applicable to ....

Y.G. Leclerc, "Capturing the local structure of image discontinuities in two dimensions, " in Proc. IEEE Computer Society Conference on Computer Vision and Pattern Recognition, San Francisco, CA, June, 1985, pp. 34--38.

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