Showing papers on "Corner detection published in 1991"

Adaptive smoothing: a general tool for early vision

Abstract: A method to smooth a signal while preserving discontinuities is presented. This is achieved by repeatedly convolving the signal with a very small averaging mask weighted by a measure of the signal continuity at each point. Edge detection can be performed after a few iterations, and features extracted from the smoothed signal are correctly localized (hence, no tracking is needed). This last property allows the derivation of a scale-space representation of a signal using the adaptive smoothing parameter k as the scale dimension. The relation of this process to anisotropic diffusion is shown. A scheme to preserve higher-order discontinuities and results on range images is proposed. Different implementations of adaptive smoothing are presented, first on a serial machine, for which a multigrid algorithm is proposed to speed up the smoothing effect, then on a single instruction multiple data (SIMD) parallel machine such as the Connection Machine. Various applications of adaptive smoothing such as edge detection, range image feature extraction, corner detection, and stereo matching are discussed. >

Early jump-out corner detectors

Abstract: Two corner detectors are presented, one of which works by testing similarity of image patches along the contour direction to detect curves in the image contour, and the other of which uses direct estimation image curvature along the contour direction. The operators are fast, robust to noise, and self-thresholding. An interpretation of the Kitchen-Rosenfeld corner operator is presented which shows that this operator can also be viewed as the second derivative of the image function along the edge direction. >

Corner-based image matching

Abstract: An image matching method that uses corners of objects, rather than their edges, to determine if the objects match. Corner locations are obtained by finding changes of curvature in the edge boundaries of the image. The corners of a reference image are used to create a distance array, which is used as a look-up table to determine how far each corner of a test array is from a corresponding corner of the reference array. A close corner-to-corner relationship indicates a good match.

Experimental methodology for performance characterization of a line detection algorithm

Abstract: We present a general methodology for designing experiments to quantitatively characterize lowlevel computer vision algorithms. The methodology can be applied to any vision problem that can be posed as a detection task. It provides a convenient framework to measure the sensitivity of an algorithm to various factors that affect the performance. The methodology is illustrated by applying it to a line detection algorithm consisting of the second directional derivative edge detector followed by a Hough transform. In particular we measure the selectivity of the algorithm in the presence of an interfering oriented grating and additive Gaussian noise. The final result is a measure of the detectors'' performance as a function of the orientation of the interfering grating.© (1991) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

The dissimilarity corner detector

Abstract: The authors present a corner detection that works by using dissimilarity along the contour direction to detect curves in the image contour. The operator is fast, robust to noise and almost self-thresholding. The standard deviation of the image noise must be specified, but this value is easily measured and the explicit modeling of image noise contributes to the robustness of the operator to noise. The authors also present a new interpretation of the Kitchen-Rosenfeld corner operator (1982) in which they show that this operator can also be viewed as the second derivative of the image function along the edge direction. >

Robot self-location based on corner detection

Abstract: This paper presents a decision theoretic method of establishing the position of a mobile robot in a known environment. Boundaries of regions of varying light intensities may be extracted from visual data gathered by the rotation of a camera about the robot position. Some of these boundaries will represent corners of the region. The identification of these corners may further be enhanced using range data. The methods in this paper rely on the probability of viewing corners and on the probability density functions of the measured view angle of corners or the separation angle between corners. View angles are used when compass knowledge is available otherwise corner separation angles are used. The probability density functions of these corner angles are derived from the region geometry and prior knowledge (if any) of the robot position. The known environment is decomposed into visibility regions for sets of corners. The probability of viewing a set of corners depends on the likelihood of the robot being positioned in one of these visibility regions. An optimal decision procedure is established to identify the corner set (or visibility region) based on visual data from a circular scan about the robot position. With this information a least squares estimate of the robot position is derived.© (1991) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

A VLSI architecture for a half-edge-based corner detector

Abstract: Corner detection is a low-level feature detection operator that is of great use in image processing applications, for example, optical flow and structure from motion by image correspondence. The detection of corners is a computationally intensive operation. Past implementations of corner detection techniques have been restricted to software. In this paper we propose an efficient very large-scale integration (VLSI) architecture for detection of corners in images. The corner detection technique is based on the half-edge concept and the first directional derivative of Gaussian. Apart from the location of the corner points, the algorithm also computes the corner orientation and the corner angle and outputs the edge map of the image. The symmetrical properties of the masks are utilized to reduce the number of convolutions effectively, from eight to two. Therefore, the number of multiplications required per pixel is reduced from 1800 to 392. Thus, the proposed architecture yields a speed-up factor of 4.6 over conventional convolution architectures. The architecture uses the principles of pipelining and parallelism and can be implemented in VLSI.

Image pick-up system with optical image shape conversion

Abstract: In an imaging system (2), comprising an image pick-up device with an image sensor (13) having a detection face (15) which is subdivided into discrete detection sub-faces for converting a radiation intensity distribution on the detection face into an electric signal, the detection sub-faces being arranged in a matrix of n rows and p columns, an object plane (5) and an optical sysstem (7, 9, 11) which cooperates with the object plane and the image sensor in order to image the object plane on the detection face, and also comprising display means (21) which are connected to the image sensor in order to display the radiation intensity distribution on the detection face, the optical system imaging a circle situated in the object plane as an ellipse in the case of a rectangular detection face having a long side and a short side, the optical system is adjusted so that the axes of the ellipse are shorter than the sides of the detection face so that video images can be formed at 100 Hz. In the case of non-square detection sub-faces, the ratio of the axes of the ellipse can be adapted in order to image a circular image on an equal number of rows and columns of detection sub-faces. The effective dimension of the detection sub-faces thus becomes square. When several, mutually shifted image sensors (13a, b) are used, the optical system is capable of performing an adaptation in order to correct the decrease of the effective dimension of the detection sub-faces. A CCD sensor suitable for use in an imaging system operating according to the PAL standard can be made suitable for use in an imaging system operating according to the NTSC standard by utilizing correct image compression.

The Inference of Structure in Images using Multi-local Quadrature Filters

Abstract: Two techniques are presented for corner detection. First, a band of filters are applied with equal radial spatial frequency, but different orientation preferences locally in the image domain. From the energy response, a linear Fourier transform is taken to give confidence measures of both “cornerness” and “edgeness. Second, we consider a multi-local spatial separation of filters that lie on a constant radius from a point of interest. This second stage of processing allows a wider classification of image structure. As a result, we infer the presence of line end points, “L”, “T”, “Y” and “X” junctions using epistemic probabilities. The results are indicative of a relationship between Fourier and Spatial domain models of filtering.

Moving edges detection in time-varying image sequences

Abstract: A new computational paradigm for the detection of moving edges in time-varying image sequences is presented. The frames need not be contiguous. The detector includes both motion and edge detection in a way that improves overall performance. >

A Hypothesize and Test Technique for Corner Detection

Abstract: In this paper a new technique is proposed for detection of corners in images of polygonal objects. The corner detector works on the principle of ‘hypothesize and test’. Initial hypotheses about the...