Showing papers on "Corner detection published in 2010"

Faster and Better: A Machine Learning Approach to Corner Detection

Abstract: The repeatability and efficiency of a corner detector determines how likely it is to be useful in a real-world application. The repeatability is important because the same scene viewed from different positions should yield features which correspond to the same real-world 3D locations. The efficiency is important because this determines whether the detector combined with further processing can operate at frame rate. Three advances are described in this paper. First, we present a new heuristic for feature detection and, using machine learning, we derive a feature detector from this which can fully process live PAL video using less than 5 percent of the available processing time. By comparison, most other detectors cannot even operate at frame rate (Harris detector 115 percent, SIFT 195 percent). Second, we generalize the detector, allowing it to be optimized for repeatability, with little loss of efficiency. Third, we carry out a rigorous comparison of corner detectors based on the above repeatability criterion applied to 3D scenes. We show that, despite being principally constructed for speed, on these stringent tests, our heuristic detector significantly outperforms existing feature detectors. Finally, the comparison demonstrates that using machine learning produces significant improvements in repeatability, yielding a detector that is both very fast and of very high quality.

Adaptive and generic corner detection based on the accelerated segment test

Abstract: The efficient detection of interesting features is a crucial step for various tasks in Computer Vision. Corners are favored cues due to their two dimensional constraint and fast algorithms to detect them. Recently, a novel corner detection approach, FAST, has been presentedwhich outperforms previous algorithms in both computational performance and repeatability. We will show how the accelerated segment test, which underlies FAST, can be significantly improved by making it more generic while increasing its performance.We do so by finding the optimal decision tree in an extended configuration space, and demonstrating how specialized trees can be combined to yield an adaptive and generic accelerated segment test. The resulting method provides high performance for arbitrary environments and so unlike FAST does not have to be adapted to a specific scene structure. We will also discuss how different test patterns affect the corner response of the accelerated segment test.

Image registration based on corner detection and affine transformation

Abstract: Image registration is an important part of the image processing and computer vision. On the basis of analyzing two types of image registration, an automated image registration method was put forward to dealing with image registration with similar transformation. At first, Corner feature was extracted by the improved multi-scale Harris operator, next, the image edge detection was conducted by canny operator, corner's coarse matching may be realized based on correlation coefficient between the corner neighborhood on the edge image, and then fine matching can be achieved, two pairs of optimal matching corner were selected from matched corners as control points of affine transformation, thus, affine transformation model could be obtained, and the registered image was performed affine transformation in order to realizing image registration. Theoretically speaking, this method's advantage isn't subjected to change limitation of the displacement, rotation and scale between two images. Finally, the simulation experiment was performed to verify and analyze the algorithm characteristics; the results show that the automatic registration algorithm is correct and effective.

Multilayered Image Processing for Multiscale Harris Corner Detection in Digital Realization

Abstract: The PC-based software programming used in complex or luxuriant image processing algorithms is time consuming and resource wasting. As appropriate processing for the image data indeed speedups complicated algorithms, we focus on a crucial case - multilayered processes. In this paper, we gauge deeply into the data flow of multilayered image processing to avoid waiting for the result from every previous steps to access the memory which occurs in many applicable algorithms. Based on combining the parallel and pipelined properties to eliminate unnecessary delays, we propose new visual pipeline architecture and use field programmable gate array to implement our hardware scheme. For verification, the multiscale Harris corner detector in cooperating with shape context and thin-plate splines were combined to complete our real-time experiment of the integrated hardware and software (H/S) system for pattern recognition.

A Spatial-Feature-Enhanced MMI Algorithm for Multimodal Airborne Image Registration

Abstract: Maximization of mutual information (MI) (MMI) has a wide application in multimodal image registration. The conventional MMI algorithm is based on the statistics derived from image sampling and does not use spatial features. In this paper, we proposed an improved MMI registration algorithm that combines the spatial information through a feature-based selection mechanism. A Harris corner filter is chosen to classify the pixels. The classification results are used to focus sample selection, joint entropy estimation, and weighted MI information calculation. Although features play an important role, conventional feature matching is not used, which removes an important challenge to the use of feature-based information. The advantages of the proposed algorithm are improved robustness by incorporation of spatial features and significant reduction in computing time for registration. Experiments with challenging synthetic images and multimodal airborne infrared images are provided to demonstrate the improvement.

Vision-based bicyclist detection and tracking for intelligent vehicles

Abstract: This paper presents a vision-based framework for intelligent vehicles to detect and track people riding bicycles in urban traffic environments. To deal with dramatic appearance changes of a bicycle according to different viewpoints as well as nonrigid nature of human appearance, a method is proposed which employs complementary detection and tracking algorithms. In the detection phase, we use multiple view-based detectors: frontal, rear, and right/left side view. For each view detector, a linear Support Vector Machine (SVM) is used for object classification in combination with Histograms of Oriented Gradients (HOG) which is one of the most discriminative features. Furthermore, a real-time enhancement for the detection process is implemented using the Integral Histogram method and a coarse-to-fine cascade approach. Tracking phase is performed by a multiple patch-based Lucas-Kanade tracker. We first run the Harris corner detector over the bounding box which is the result of our detector. Each of the corner points can be a good feature to track and, in consequence, becomes a template of each instance of multiple Lucas-Kanade trackers. To manage the set of patches efficiently, a novel method based on spectral clustering algorithm is proposed. Quantitative experiments have been conducted to show the effectiveness of each component of the proposed framework.

The Improbability of Harris Interest Points

Abstract: An elementary characterization of the map underlying Harris corners, also known as Harris interest points or key points, is provided. Two principal and basic assumptions made are: (1) Local image structure is captured in an uncommitted way, simply using weighted raw image values around every image location to describe the local image information, and (2) the lower the probability of observing the image structure present in a particular point, the more salient, or interesting, this position is, i.e., saliency is related to how uncommon it is to see a certain image structure, how surprising it is. Through the latter assumption, the axiomatization proposed makes a sound link between image saliency in computer vision on the one hand and, on the other, computational models of preattentive human visual perception, where exactly the same definition of saliency has been proposed. Because of this link, the characterization provides a compelling case in favor of Harris interest points over other approaches.

Automated detection of blob structures by Hessian analysis and object scale

Abstract: Automated detection of blob-like structures is desirable in many biomedical applications such as nodule detection in radiographs and CT images, lymph nodes detection in CT images, and cell counting or tracking in biological images. Multiscale analysis of Hessian matrix is widely used for enhancement or detection of blob-like structures in two-dimensional (2D) and three-dimensional (3D) images. We proposed a new blob detector and a new detection response measure, blobness, based on eigenvalues of the Hessian matrix and local object scale. Pixels with higher blobness are clustered as detected blobs. We evaluated our method by comparison with two existing methods on both simulated and real images. Our results indicated that our automated blob detector had better performance on those images especially when the blobs were close to each other. Our method can be easily extended to 3D for computer-aided detection of blob-like structures in medical images.

Iris recognition using corner detection

Abstract: This paper proposes an iris recognition algorithm with the help of corner detection. It consists of five major steps i.e., iris acquisition, localization, normalization, feature extraction and matching. The inner pupil boundary is localized using Circular Hough Transformation. The technique performs better in the case of occlusions and images muddled by artifacts such as shadows and noise. The outer iris boundary is detected by circular summation of intensity approach from the determined pupil center and radius. The localized iris image is transformed from Cartesian to polar co-ordinate system to handle different size, variation in illumination and pupil dilation. Corners in the transformed iris image are detected using covariance matrix of change in intensity along rows and columns. All detected corners are considered as features of the iris image. For recognition through iris, corners of both the iris images are detected and total number of corners that are matched between the two images are obtained. The two iris images belong to the same person if the number of matched corners is greater than some threshold value. The system is tested on a database having 900 iris images and also on CASIA database. The algorithm has shown an overall accuracy of 95.4% with FRR of 5% and FAR of 4%.

Farsi/Arabic text extraction from video images by corner detection

Abstract: Video text information plays an important role in semantic-based video analysis, indexing and retrieval. In this paper, we proposed a novel Farsi text detection approach based on intrinsic characteristics of Farsi text lines, which is more robust to complex backgrounds and various font styles. First, by an edge detector operator, all the possible edges in vertical, horizontal, 45 and 135 degrees are extracted. Then, for extracting text strokes, some pre-processing such as dilation and erosion are done according to the font size. Afterward, by finding the edges cross points, corners map is extracted. To discard non-text corners and finding real font size, histogram analysis is done. After finding real font size, input image is rescaled and a new corner map is extracted. Finally, the detected candidate text areas undergo the empirical rules analysis to identify text areas and project profile analysis for verification and text lines extraction. Experimental results demonstrate that the proposed method is robust to font size, font colour, and background complexity.

Three-dimensional human face reconstruction method based on stereoscopic vision

Abstract: The invention relates to a three-dimensional human face reconstruction method based on stereoscopic vision of two eyes. Steps for implementing the method are as follows: (1) using two cameras to respectively take a picture of a human face from two different angles; (2) respectively determining an internal parameter matrix and an external parameter matrix of each camera; (3) carrying out extreme line correction and picture conversion according to the determined data; (4) utilizing a Harris corner detection operator to extract characteristics of the human face, and carrying out initial matching by utilizing a local template window method and extreme line restriction; (5) starting from an initial matching assemblage, and obtaining a sparse matching assemblage by utilizing a seed point growing algorithm; (6) taking the sparse matching assemblage as a guiding point and carrying out a dynamic planning algorithm so as to finish dense matching; and (7) calculating three-dimensional coordinates of actual physical points on the human face according to the determined data and the matching relationship, thereby reconstructing a three-dimensional point cloud of the human face.

Effective Image Registration based on Improved Harris Corner Detection

Abstract: Feature selection is a key step for image registration. The success of feature selection has a fundamental effect on matching image. Corners determine the contours characteristics of the target image, and the number of corners is far smaller than the number of image pixels, thus can be a good feature for image registration. By considering the algorithm speed and registration accuracy of the image registration, the paper proposes an improved Harris corner detection method for effective image registration. This method effectively avoids corner clustering phenomenon occurs during the corner detection process, thus the corner points detected distribute more reasonably, and the image registration become faster. The experiments also showed the effect of image registration is satisfactory, and reaches a reasonable match.

Electronic image stabilizing method for digital videos

Abstract: The invention relates to an electronic image stabilizing method for digital videos, belonging to the technical field of digital image processing. The method comprises the following steps of: firstly, extracting characteristic points in two adjacent images in a video sequence by utilizing an SUSAN corner detection operator, matching the characteristic points by utilizing a relevant matching method and rejecting false matching point pairs; then selecting three pairs from the correct matching point pairs and solving for an affine transformation parameter between two images by utilizing a corresponding coordinate relation of the matching point pairs on a two-dimensional image; and finally, carrying out an affine inverse transformation on a next image according to an affine formula to obtain an image registered with a previous image in space so as to realize the electronic image stabilization of the digital videos. The electronic image stabilizing method of the digital videos can automatically eliminate jitters of the video images in real time, wherein the jitters comprise translations and rotating movements among images.

An algorithm for vehicle detection and tracking

Abstract: In this paper, we propose a vehicle detection and tracking algorithm. The detection is done using the median filtering and blob extraction. Median filtering is used for background extraction which is later subtracted from the motion frames for object detection. Morphological operators are employed for blob extraction. Hence, object detection is achieved using median filtering and morphological closing operation. Kalman filtering is used for object tracking which uses location of blobs. One of the advantages of this system is that each vehicle in the frame is classified into different color boxes. We present preliminary research results that will finally lead to the identification of the tracked vehicle.

An improved SUSAN corner detection algorithm based on adaptive threshold

Abstract: The SUSAN operator needs to adjust similarity threshold manually time after time in order to achieve a good corner detection results, and it can't detect some corners with special or complex shape. In view of this, the SUSAN operator is improved by the authors of this paper. Firstly, an adaptive threshold selection method based on iterative operation is proposed, and then a discrete ring-shaped mask is added within the SUSAN's circular mask so as to overcome the deficiency of the SUSAN operator. The experimental results show that the improved SUSAN algorithm can not only greatly improve the automation of program operation by releasing the programmers of adjusting threshold manually, but also obtain a good detection result for various types of corners.

Hierarchical filtered motion field for action recognition

Abstract: Described is a hierarchical filtered motion field technology such as for use in recognizing actions in videos with crowded backgrounds. Interest points are detected, e.g., as 2D Harris corners with recent motion, e.g. locations with high intensities in a motion history image (MHI). A global spatial motion smoothing filter is applied to the gradients of MHI to eliminate low intensity corners that are likely isolated, unreliable or noisy motions. At each remaining interest point, a local motion field filter is applied to the smoothed gradients by computing a structure proximity between sets of pixels in the local region and the interest point. The motion at a pixel/pixel set is enhanced or weakened based on its structure proximity with the interest point (nearer pixels are enhanced).

A novel zero-watermark copyright authentication scheme based on lifting wavelet and Harris corner detection

Abstract: In this paper, a novel zero-watermark copyright authentication scheme based on Internet public certification system is proposed. This approach utilizes Haar integer wavelet transform based on a lifting scheme and adaptive Harris corner detection to extract image features, which will be used to produce a binary feature map, and the map is very crucial to the generation of watermark registered later. By properly choosing the parameters of aforementioned techniques such as the threshold T and the radius of local feature region R, the feature map is so much more stable and distinguishing that it can be used to construct robust watermark. Simulations demonstrate that the proposed scheme is resistant to many kinds of signal processing and malicious attacks such as Gaussian blurring, additive noising, JPEG lossy compression, cropping, scaling and slight rotation operation. Compared with a relative scheme such as that of Chang’s, the scheme in this paper is more practicable and reliable and can be applied to the area of copyright protection.

Object recognition using junctions

Abstract: In this paper, we propose an object detection/recognition algorithm based on a new set of shape-driven features and morphological operators. Each object class is modeled by the corner points (junctions) on its contour. We design two types of shape-context like features between the corner points, which are efficient to compute and effective in capturing the underlying shape deformation. In the testing stage, we use a recently proposed junction detection algorithm [1] to detect corner points/junctions on natural images. The detection and recognition of an object are then done by matching learned shape features to those in the input image with an efficient search strategy. The proposed system is robust to a certain degree of scale change and we obtained encouraging results on the ETHZ dataset. Our algorithm also has advantages of recognizing object parts and dealing with occlusions.

Log-log scaled Harris corner detector

Abstract: The well-known Harris algorithm for corner detection in digital images suffers from poor localisation especially in high (≥3) order corners. In this reported work, the Harris algorithm is extended to have better localisation performance by applying the log-log scale to gradient vectors of image pixels.

Image retrieval using Contourlet based interest points

Abstract: In this work we present the method for image retrieval based on the Non-Subsampled Contourlet Transform (NSCT) and the Harris corner detector. The NTSC-based interest point detector is proposed by combination of NTSC and Harris corner detector called the Contourlet Harris detector. We also present the method how to extract the image features using this Contourlet Harris detector that is applied for image retrieval. Experiments are implemented on the WANG database aiming to compare retrieval effectiveness of proposed method to some methods have announced. Results demonstrate that the proposed method shows a quite improvement in the retrieval effectiveness.

A New Technique for Solving Puzzles

Abstract: This paper proposes a new technique for solving jigsaw puzzles. The novelty of the proposed technique is that it provides an automatic jigsaw puzzle solution without any initial restriction about the shape of pieces, the number of neighbor pieces, etc. The proposed technique uses both curve- and color-matching similarity features. A recurrent procedure is applied, which compares and merges puzzle pieces in pairs, until the original puzzle image is reformed. Geometrical and color features are extracted on the characteristic points (CPs) of the puzzle pieces. CPs, which can be considered as high curvature points, are detected by a rotationally invariant corner detection algorithm. The features which are associated with color are provided by applying a color reduction technique using the Kohonen self-organized feature map. Finally, a postprocessing stage checks and corrects the relative position between puzzle pieces to improve the quality of the resulting image. Experimental results prove the efficiency of the proposed technique, which can be further extended to deal with even more complex jigsaw puzzle problems.

3D ultrasound volume stitching using phase symmetry and harris corner detection for orthopaedic applications

Abstract: Stitching of volumes obtained from three dimensional (3D) ultrasound (US) scanners improves visualization of anatomy in many clinical applications. Fast but accurate volume registration remains the key challenge in this area.We propose a volume stitching method based on efficient registration of 3D US volumes obtained from a tracked US probe. Since the volumes, after adjusting for probe motion, are coarsely registered, we obtain salient correspondence points in the central slices of these volumes. This is done by first removing artifacts in the US slices using intensity invariant local phase image processing and then applying the Harris Corner detection algorithm. Fast sub-volume registration on a small neighborhood around the points then gives fast, accurate 3D registration parameters. The method has been tested on 3D US scans of phantom and real human radius and pelvis bones and a phantom human fetus. The method has also been compared to volumetric registration, as well as feature based registration using 3D-SIFT. Quantitative results show average post-registration error of 0.33mm which is comparable to volumetric registration accuracy (0.31mm) and much better than 3D-SIFT based registration which failed to register the volumes. The proposed method was also much faster than volumetric registration (~4.5 seconds versus 83 seconds).

Study on panoramic image stitching algorithm

Abstract: This paper has a thorough study on panoramic image, an important application of virtual reality technology, and proposes a feature based method of generating a panorama Descriptions of solutions and algorithms for each step were given in details, including Harris corner detection algorithm, RANSAC algorithm, Levenberg-Marquardt algorithm and so on Some improvements of the algorithm were proposed in order to resolve the deficiencies The experimental results were given at the end of the paper It shows that the panoramic generated by this method is complete, clear, and no obvious distortion, and the methods adopted in the system are appropriate and practical

Auto-adaptive harris corner detection algorithm based on block processing

Abstract: To eliminate the problems of extracting false corners and losing information of real corners and overcome the difficulty in finding a universal threshold in the non-maximal inhibition for the processing of all pictures in the Harris corner detection algorithm, an auto-adaptive threshold is introduced in this paper in order to generate more accurate corners. In addition, a method of block processing to divide an image into several blocks and process each block independently is proposed to ensure that the corners detected are evenly distributed in the image without clustering and thus eliminate the possibility that some corners may be lost because of the sharp contrast in gray scale in different parts of the image. Experimental results showed that this improved algorithm outperformed traditional and previous methods both in accuracy and evenness of distribution of detected corners.

High Definition Feature Map for GVF Snake by Using Harris Function

Abstract: In image segmentation the gradient vector flow snake model is widely used. For concave curvatures snake model has good convergence capabilities, but poor contrast or saddle corner points may result in a loss of contour. We have introduced a new external force component and an optimal initial border, approaching the final boundary as close as possible. We apply keypoints defined by corner functions and their corresponding scale to outline the envelope around the object. The Gradient Vector Flow (GVF) field is generated by the eigenvalues of Harris matrix and/or the scale of the feature point. The GVF field is featured by new functions characterizing the edginess and cornerness in one function. We have shown that the max(0,log[max(λ 1, λ 2)]) function fulfills the requirements for any active contour definitions in case of difficult shapes and background conditions. This new GVF field has several advantages: smooth transitions are robustly taken into account, while sharp corners and contour scragginess can be perfectly detected.

Points of interest detection in cervical spine radiographs by polygonal approximation

Abstract: In this paper, we introduce a robust approach to detect points of interest in cervical spine radiographs. The perspective of this work is to segment the vertebrae on X-Ray images for the analysis of the vertebral mobility. In previous work, we proposed a segmentation technique based on Active Shape Model. The extraction and the detection of the vertebra corners can contribute to the automatic initialization of the Active Shape Model search and can give valuable information about the spine curvature. Here, we present the benefits of the polygonal approximation dedicated to the points of interest detection. The methodology developed here is composed of 3 stages: a contrast limited adaptive histogram equalization, a Canny edge detection filter and an edge polygonal approximation. The first histogram equalization step is a pretraitment needed to improve the image quality in order to perform a better contour detection. The Canny operator detects the edges in the radiograph which are used as an input to the polygonal approximation. The edges become segment lines whose intersections define corners. We compare the results obtained with our approach based on the polygonal approximation to results coming from the Harris corner detector.

Position estimation of autonomous aerial navigation based on Hough transform and Harris corners detection

Abstract: Automatic position estimation of Unmanned Aerial Vehicle remains to be one of the challenging problems in computer vision. When an Inertial Navigation System (INS) is inadequate and its position errors compounds over time, the main objective of this work approach is to investigate efficient algorithms for localization estimation for Autonomous Aerial Navigation. Thus, our approach is based on line detection by Hough transform and Harris corners detection. This method consists of firstly to detect the feature extraction using Hough transform in the images and estimate the rotation. Secondly the method consists of scale estimation based in distribution of features point in image. Thirdly we estimate the translation parameter. Finally we use a geometric model estimation to map the high-resolution image onto a low-resolution image, which results in position estimation.

Robust Detection of Corners and Corner-line Links in Images

Abstract: We define corner points in an image as the intersections among detected straight line segments, and propose an algorithm that detects corners from such a definition. Our corner detection algorithm CLDC then makes use of the LDC (Line Detection using Contours) algorithm from [19], which outputs the list of all detected line segments together with their endpoints. Each line segment is extended in a post-processing step. CLDC (Corners from LDC) then finds corners in O((n+I)log n) time, where n and I are the number of endpoints the intersections of line segments, respectively. Detected corners are linked via line segments that define them. Such an output of the corner detection algorithm is a novel concept. The algorithm is comparable in time complexity with other algorithms, while providing more information about the line segments in the image. CLDC is robust to image transformations, such as rotation and translations. Our CLDC is compared to some existing algorithm, and its advantages are demonstrated.