next up previous contents
Next: The Local Character of Up: Computer Vision Previous: Visual Shape of Curve

Scale Space Surfaces and the Computation of Space Curves from Focal and Penumbral Blur

J. Elder, S.W. Zucker We address the problem of how to reason within scale-space to produce reliable estimates of image structure. Our approach depends on the observation that while the structure of the scene cannot be known in advance, the parameters of the sensing apparatus are known, and can be used to define unique surfaces in scale space based on the notion of a locally-defined minimum reliable scale. We demonstrate the usefulness of these surfaces by applying them to the problem of detecting edges from images with shallow depth of field and shadows. We show that edges spanning a broad range of blur scale can be recovered accurately by a single system with no input parameters other than the second moment of the sensor noise. A natural dividend of this approach is a reliable measure of the thickness of contours which can be used to estimate depth from defocus along image contours (Fig. 11). [branches-segmentation.eps].45 A novel method for blur estimation in which the scale of estimation is adapted to the local image structure allows depth segmentation in complex images where conventional Fourier-based techniques would fail.zucker:branches-segmentation



Thierry Baron
Mon Apr 7 12:54:24 EDT 1997