Speeded Up Robust Features Detector Research Articles

Comparative Analysis for Bag of Features (BoF) Performance

The accuracy of the Bag of Features (BoF) is greatly affected by the discriminatory power of feature extraction techniques. This paper presented a proposed method designed to find the best features technique for constructing a BoF model according to the image classification. It consists of four stages: feature extraction, where detectors and descriptor feature techniques have been exploited to generate different BoF models. Each BoF model is generated depending on what detector and descriptor are used. The BoF models are constructed to represent the images as feature vectors. The classification process is then performed on two image datasets. Finally, the efficiency of BoF models is analyzed and evaluated with respect to the accuracy of their classification performance. Experimental results indicated that the best level of accuracy was provided by the proposed BoF model with the KAZE features method. The results also showed that the BoF model with Speeded Up Robust Features (SURF) was superior to other feature methods in terms of execution time, which was 0.01218 seconds. Moreover, the BoF model generated by the SURF detector combined with the KAZE descriptor achieved a high level of accuracy of 0.99 and kept the time complexity low (0.01948 seconds).

Multi-scale super-pixels-based passive forensics for copy-move forgery using local characteristics

Copy-move forgery is the most common kind of tampering technique for digital images. This paper presents a novel hybrid approach, which uses the Speeded-Up Robust Features (SURF) point and characteristics of Local Feature Regions (LFRs) matching. First, the multi-scale super-pixels algorithm adaptively divides the suspect image into irregular blocks according to the texture level of host images. Then, the improved SURF detector is adopted in extracting feature points from each super-pixel and the feature point threshold is related to the entropy of each super-pixel block. Next, LFRs are defined, and a robust feature descriptor is extracted from each LFR as a vector field. Last, the matching LFRs are found by using Euclidean locality sensitive hashing; the removal of falsely matched pairs is realised by using the Random Sample Consensus (RANSAC) algorithm. Comparing with the leading-edge block-matching methods and point-based methods, our method can produce far better detection results.

A Front Water Recognition Method Based on Image Data for Off-Road Intelligent Vehicle

Off-road intelligent vehicle is an important application about Internet of Vehicles technology used in the transportation field, and the front obstacle recognition method is the key technology for off-road intelligent vehicle. In this paper, based on smart data aggregation inspired paradigm of IoT applications, we mainly study perception technology in vehicle networking by using image data and one symmetrical speeded-up robust features detector (SURF). By considering symmetry and image data aggregation, we found that data aggregation had the ability of providing global information for Internet of Vehicles systems. After we have built the experiment platform, the experiment results showed that this method is faster than Scale-Invariant Feature Transform algorithm in this case, which can satisfy the water detection accuracy and the real-time requirement. So, this method is effective for the water images detection with great symmetry to off-road intelligent vehicle, and it also gives a useful reference about environment perception technology and smart data aggregation inspired paradigm used in future Internet of Vehicles, intelligent vehicle, and traffic safety applications.

Ground Control Point Automatic Extraction for Spaceborne Georeferencing Based on FPGA

Feature points that are obtained from the combined speeded-up robust feature (SURF) detector and binary robust independent elementary features (BRIEF) descriptor have a highly robust performance. These points are previously considered the ground control points (GCPs) for building a connection between the image coordinates and the corresponding geodetic coordinates. This article proposes a novel architecture to automatically and intelligently extract GCPs based on field programmable gate arrays (FPGAs). The parallelization SURF detector, BRIEF descriptor, and BRIEF matching are implemented in a single Xilinx XC7VX980T FPGA system. Word length reduction, memory-efficient parallel architecture, shift and subtraction strategies, a sliding window for separable convolution, and an optimized multispacer-scale are used to optimize the SURF detector. Improved parallel adder trees are used to accelerate the BRIEF matching. The proposed system achieves 380 frame per second (fps) with a 100 MHz clock frequency, which satisfies the real-time and low-power requirements of embedded devices. The results of the experiment demonstrate that the proposed architecture, when mapped onto a Xilinx Virtex-7 XC7VX980T FPGA device, can select the robust feature points.

Unsupervised feature learning for environmental sound classification using Weighted Cycle-Consistent Generative Adversarial Network

In this paper we propose a novel environmental sound classification approach incorporating unsupervised feature learning via the spherical K-Means++ algorithm and a new architecture for high-level data augmentation. The audio signal is transformed into a 2D representation using a discrete wavelet transform (DWT). The DWT spectrograms are then augmented by a novel architecture for cycle-consistent generative adversarial network. This high-level augmentation bootstraps generated spectrograms in both intra-and inter-class manners by translating structural features from sample to sample. A codebook is built by coding the DWT spectrograms with the speeded-up robust feature detector and the K-Means++ algorithm. The Random forest is the final learning algorithm which learns the environmental sound classification task from the code vectors. Experimental results in four benchmarking environmental sound datasets (ESC-10, ESC-50, UrbanSound8k, and DCASE-2017) have shown that the proposed classification approach outperforms most of the state-of-the-art classifiers, including convolutional neural networks such as AlexNet and GoogLeNet, improving the classification rate between 3.51% and 14.34%, depending on the dataset.

High-resolution image registration based on improved SURF detector and localized GTM

ABSTRACTHigh-resolution image registration is an important task in remote sensing image processing. In this paper, an automatic and robust local feature-based image registration approach is proposed for high-resolution remote sensing images. The proposed method consists of four main steps. In the first step, an integrated local feature-based matching method based on an improved speeded-up robust features (SURF) detector and an adaptive binning scale-invariant feature transform (AB-SIFT) descriptor is developed for fast, dense and robust tie-point extraction. In the second step, a localized graph transformation matching (LGTM) method is developed for reliable mismatch elimination. In the third step, an advanced oriented least square matching (OLSM) method is applied to improve the positional accuracy of the refined tie-points. Finally, the input image is warped using an appropriate transformation model. To investigate the impact of the transformation function, the capability of some models, including, polynomials of degrees 2 to 4, piecewise linear (PL), weighted mean (WM) and multiquadric (MQ) are compared. The proposed method has been evaluated with five pairs of high-resolution remote sensing images from IRS-P5, SPOT 5, SPOT 6, IKONOS, Geoeye, Quickbird, and Worldview sensors, and the registration results demonstrate its robustness and capability. The MATLAB code of the proposed method can be downloaded from https://www.researchgate.net/publication/320354469_HRImReg.

Propose retina identification system based on the combination of SURF detector and BRISK descriptor

In this paper the design of hybrid retina matching algorithm that is used in identification systems is considered. Retina based recognition is apparent as the most secure method for identification of an identity utilized to differentiate persons. The characteristics of Speeded up Robust Feature (SURF) and Binary Robust Invariant Scalable Key-Points (BRISK) algorithm have been used in order to produce a fast matching algorithm than the classical ones, those characteristics are important for real-time applications which usually need quick processing of a growing quantity of data. The algorithm is divided into three stages: retinal image processing and segmentation, extracting the local features from the retinal images using SURF and BRISK descriptors and finally matching those features using KNN matcher. The algorithm was trained with structured analysis of retina (STARE) database giving accuracy that reach 100%.

A new keypoint-based copy-move forgery detection for color image

Over the past decade, many efforts have been made in copy-move forgery detection (CMFD), and some promising methodologies have been proposed to detect copy-move forgeries. Keypoint based CMFD approaches extract image interest points and use local visual features to identify duplicated regions, which exhibit remarkable performance with respect to memory requirement and computational cost. But unfortunately, they usually use the pure gray-based detectors to extract interest points in which the significant color information is ignored. Also, local visual features are computed without considering the correlation between different color channels. All this lower inevitably the detection and localization accuracy for color tampered image. In this paper we propose a new technique for the detection and localization of copy-move forgeries, which is based on color invariance model and quaternion polar complex exponential transform (QPCET). First, stable color image interest points are extracted by using new interest point detector, in which the SURF (speeded up robust features) detector and color invariance model are incorporated. Then, a set of connected Delaunay triangles is built based on the extracted color image interest points, and suitable local visual features of the triangle mesh are computed using QPCET. Afterwards, local visual features are employed to match triangular meshes by a combination of reversed-generalized 2 nearest-neighbor (Rg2NN) and best bin first (BBF). Finally, the falsely matched triangular meshes are removed by customizing the random sample consensus, and the duplicated regions are localized using zero mean normalized cross-correlation measure. Compared with the state-of-the-art approaches, extensive experimental results prove that our proposed method can detect and localize color image copy-moves with good accuracy even in adverse conditions.

Sistem Pentautan Citra Udara Menggunakan Algoritme SURF dan Metode Reduksi Data

One of the algorithm for aerial image stitching system is SURF (Speeded Up Robust Features). It is a robust algorithm which is invariant to image scale, rotation, blurring, illumination, and affine transformation. Although SURF has good performance, some of the detected keypoints are not always considered as necessary keypoints . As a result, these unnecessary keypoints are needed to be eliminated to decrease computation time.The proposed system uses SURF detector in the detection process. The data reduction method will eliminate couple of keypoints which have near distance each other. Next, the keypoints will be described by SURF descriptor.The description Results further matched using FLANN. The next step is the search pattern with RANSAC homography matrix and stitch the picture to accumulate keypoints using warpPerpective.Stitching system are tested with some variations, such as scale variations, rotation variations, and overlap variations on the image. Based on the result, the proposed Data Reduction method has optimum value of minimal radius from 40 pixels to 100 pixels. The stitching process is still working with up to 90% keypoint number reduction. Average computation time using data reduction method are 39,41% faster than without data reduction method.

Automatic Deformity Estimation for Thoracic Section Between Inhale and Exhale Positions

Corresponding control point pairs or landmarks in images can be used to define the deformation with respect to time, point of view or modality. Manual definition of the number of control points in an image, enough to define all kinds of deformation is a tedious task. Hence, automatic definition of control points is the way forward taken in this proposed work. The paper proposes an automatic registration process for tracing of the deformity path of the thoracic region based on feature detector speeded up robust feature (SURF) and moving least squares (MLS). The set of control points on the images is defined by its feature set which is obtained by using the SURF detector, which serves as input for MLS algorithm to trace the deformations of the image (thoracic image in this case). The credibility and performance of the above proposed method is demonstrated by its outstanding experimental results.

Robust digital watermarking based on local invariant radial harmonic fourier moments

It is a challenging work to design a robust image watermarking scheme against local geometric distortions. Radial harmonic Fourier moments (RHFMs) is a powerful tool in image processing owing to its better image description capability, lower noise sensitivity, and geometric invariance property. Based on SURF (Speeded-up robust features) detector and invariant RHFMs, we proposed a robust image watermarking scheme against local geometric distortions in this paper. Firstly, the SURF detector, which is a fast and highly performant multiscale feature detector, is used to extract image feature points in original image. Secondly, the stable and non-overlapped local circular regions centered at feature points are thereafter constructed and selected by combining characteristic scales and feature points refinement. And finally, the digital watermark is embedded in the invariant RHFMs of the local circular regions. Similarly to watermark insertion, the digital watermark can be blindly detected from the local circular regions. Experimental results confirm the validity of our approach and its higher robustness against various attacks (especially local geometric distortions) compared to alternative watermarking methods in the literature.

Content-based image retrieval using local visual attention feature

Content-based image retrieval (CBIR) has been an active research topic in the last decade. As one of the promising approaches, salient point based image retrieval has attracted many researchers. However, the related work is usually very time consuming, and some salient points always may not represent the most interesting subset of points for image indexing. Based on fast and performant salient point detector, and the salient point expansion, a novel content-based image retrieval using local visual attention feature is proposed in this paper. Firstly, the salient image points are extracted by using the fast and performant SURF (Speeded-Up Robust Features) detector. Then, the visually significant image points around salient points can be obtained according to the salient point expansion. Finally, the local visual attention feature of visually significant image points, including the weighted color histogram and spatial distribution entropy, are extracted, and the similarity between color images is computed by using the local visual attention feature. Experimental results, including comparisons with the state-of-the-art retrieval systems, demonstrate the effectiveness of our proposal.

Medical image registration based on SURF detector

A quick method of 2D global registration of CT images in different series of studies of one patient is suggested in this work. The distinguishing feature of this approach is the use of image registration based on the SURF (Speeded Up Robust Features) detector, which has proved efficient in computer vision.