Showing papers on "Scale-invariant feature transform published in 1994"

Comparisons of probabilistic and non-probabilistic Hough transforms

Abstract: A new and efficient version of the Hough Transform for curve detection, the Randomized Hough Transform (RHT), has been recently suggested. The RHT selects n pixels from an edge image by random sampling to solve n parameters of a curve and then accumulates only one cell in a parameter space. In this paper, the RHT is related to other recent developments of the Hough Transform by experimental tests in line detection. Hough Transform methods are divided into two categories: probabilistic and non-probablistic methods. Four novel extensions of the RHT are proposed to improve the RHT for complex and noisy images. These apply the RHT process to a limited neighborhood of edge pixels. Tests with synthetic and real-world images demonstrate the high speed and low memory usage of the new extensions, as compared both to the basic RHT and other versions of the Hough Transform.

Line Patterns Hough Transform for line segment detection

Abstract: In this paper, a modified Hough Transform (HT) for line segment extraction, namely Line Patterns Hough Transform (LPHT) is proposed. By using the concept of relative connectivity of points on a line segment, the end points of a line segment are voted in a 2-dimensional accumulator which has the same resolution as the image. Hence, the end points of the line segment can be directly extracted from the accumulator. Compared with the standard Hough Transform, this method has a shorter computing time, less memory required, higher accuracy; it is more robust and it can directly locate the end points of line segments. >

Non-analytic object recognition using the Hough transform with the matching technique

Abstract: By using the combination of the Hough and the contour sequence matching techniques, the authors suggest a new algorithm of scale and rotation invariant for pattern recognition. In the proposed algorithm, the conventional four-dimensional Hough space is replaced by a two-dimensional one. Because of the significant reduction of memory requirement, the authors can provide a much faster and more efficient generalised Hough algorithm. In order to enhance the performance of the Hough transform, they also suggest a new peak searching technique on the Hough space to achieve an accurate location of the detected objects.

Crater detection by linear filters representing the Hough Transform

Abstract: Crater detection is accomplished by means of Hough transform, applied to binary images, which is obtained by edge detection and thresholding from grey value images. Because the ordinary Hough transform is very time consuming, the problem must be reduced. Starting from minimum mean square estimation linear (matched) filters are obtained representing the radon transform, which is equivalent to the Hough transform. Because the filter coefficients can be calculated in advance and may be saved in a filter bank, the computation of the Hough accumulator array can be speeded up considerably. The computational amount can be reduced further, if the accumulator array is projected from the (a,b,r)-space to the (a,b)-space. In order to do that, a special method is presented, which uses a priori information about the possible circle (or crater) radii. Coarse estimation of the circle center coordinates by this method defines limited regions where the full Hough transform can be applied with a much lower amount of computation. The method presented here was applied to simulated images and to Mars images obtained by the VIKING Mars orbiters.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Virtual line segment-based Hough transform

Abstract: The generalized Hough transform (GHough) is a useful technique for detecting and locating 2D shapes. However, GHough requires a 4D accumulator array to detect objects of unknown scale and orientation. In this paper, we propose an extension of GHough, the virtual line segment-based Hough transform (VHough) that requires much less storage than GHough to accurately determine the scale and orientation of an object instance. VHough takes O(N/sup 2/) time, where N is the number of edge pixels in an image, but requires only 2D accumulator array for the detection of arbitrarily rotated and scaled objects. We present an experimental result to show that VHough is well-suited to recognition tasks when no a priori knowledge about parameters is available.

Object recognition with a 2-D Hough domain

Abstract: In this paper, we report a new Generalized Hough Transform (GHT) for the recognition of non-analytic objects. Our proposed GHT algorithm is able to replace the conventional 4-D Hough space with a 2-D one. The significant reduction of memory requirement leads to a great simplification of the peak searching time for the recognition. By an analysis of the angular deviation of the edge operator, we also derive a new profile for the modeling of peaks, which is extremely helpful for our peak searching and verification processes. >

A coarse-to-fine approach for circle detection

Abstract: An efficient circle detection method using a coarse-to-fine strategy is proposed. The Hough transform is performed using a rough Hough space within a small window around a feature point which is selected using local connectivity information. Hough cells which have sufficient number of votes are retained, whilst all others are discarded. These Hough cells are refined using a finer quantization and then a larger window around the feature point is employed for another Hough transform. This process is repeated till sufficient evidence for the existence of a circle is obtained. >

Novel 2-D Hough planes for the detection of ellipses

Abstract: A new approach of the Hough transform for the detection of ellipses is presented in this paper. In our algorithm, the search of a 5-D Hough domain is replaced with four 2-D Hough planes. One of the main differences between our proposed transform and other techniques is the approach to extract feature points in an image for the recognition. For the accumulation process in the Hough domain, an inherent property of our suggested algorithm is its capability to effect verification. Hence, significant improvement on the performance of the recognition is achieved. Furthermore, our proposed algorithm is applicable for the detection of both circular and elliptical objects concurrently. >

A comparison between the standard Hough transform and the Mahalanobis distance Hough transform

Abstract: The Hough Transform is a class of medium-level vision techniques generally recognised as a robust way to detect geometric features from a 2D image. This paper presents two related techniques. First, a new Hough function is proposed based on a Mahalanobis distance measure that incorporates a formal stochastic model for measurement and model noise. Thus, the effects of image and parameter space quantisation can be incorporated directly. Given a resolution of the parameter space, the method provides better results than the Standard Hough Transform (SHT), including under high geometric feature densities. Secondly, Extended Kalman Filtering is used as a further refinement process which achieves not only higher accuracy but also better performance than the SHT. The algorithms are compared with the SHT theoretically and experimentally.

Hough transform: it's application to the linearly moving point targets detection

Abstract: The Hough transform is applied to successfully detecting moving point targets in an image sequence. The behaviour and performance of the Hough transform are listed. Three realizable algorithms are provided. The complexity of the computations and noise performance are discussed. The computational simulation results are explained. >

Distance recovery of 3D objects from stereo images using Hough transform

Abstract: This paper examines recovering the depth of an object from two stereo images by correlating matching feature points. The principle concern of this paper is to demonstrate one method of feature point extraction. The orientation of the two viewers with respect to each other is fixed, this allows us to make use of the epipolar constraint for depth recovery. The methodology for extracting feature points from each image are: perform edge detection on each original image; use the Hough transform to identify the lines which define the image; label feature points as the intersections of lines where the intersection is a valid point in the edge detected image. The above method was very accurate with respect to recovering the depth of input images which were in certain orientations. However, the algorithm was too sensitive to error in the Hough transform space to allow for consistent evaluation of depth for all orientations.© (1994) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

On a Segmentation Algorithm of Lines of Hough Transform using Discriminant Analysis - on the Enforcement of the Platform of Hough Transform -.

Abstract: It is essential for edge detection of Hough transform to enforce the plaform of its availabilities. One of the most important subject of these enforcements is to prepare a robust method for segmenting line segments. We propose two kinds of algorithms: First one is a local algorithm, where a local area moves along the straight lines to judge locally whether the edge points should be on an edge segment or not. Another one is a global algorithm, where a histogram of the edge points along the straight line and the discriminant measure for it collaborate to judge globally whether the straight line should be deviated into some edge segments or not. We discussed also about the integration of these two algorithms to realize a practical line segmentation method.

An analytical least squares Hough transform

Abstract: The Hough Transform (HT) is an efficient method to extract geometric features from an image which works fairly well for images that contain noise and occlusion. However, its performance decreases with image and parameter space quantisation noise. Accuracy is limited by memory costs and oversampling effects. This paper describes two HT variants based on analytical least squares refinement based on a conventional HT voting stage. Both methods reduce memory requirements. A number of experimental results are presented and compared with the standard HT. >