Showing papers on "Edge detection published in 2014"
06 Sep 2014
TL;DR: A novel method for generating object bounding box proposals using edges is proposed, showing results that are significantly more accurate than the current state-of-the-art while being faster to compute.
Abstract: The use of object proposals is an effective recent approach for increasing the computational efficiency of object detection. We propose a novel method for generating object bounding box proposals using edges. Edges provide a sparse yet informative representation of an image. Our main observation is that the number of contours that are wholly contained in a bounding box is indicative of the likelihood of the box containing an object. We propose a simple box objectness score that measures the number of edges that exist in the box minus those that are members of contours that overlap the box’s boundary. Using efficient data structures, millions of candidate boxes can be evaluated in a fraction of a second, returning a ranked set of a few thousand top-scoring proposals. Using standard metrics, we show results that are significantly more accurate than the current state-of-the-art while being faster to compute. In particular, given just 1000 proposals we achieve over 96% object recall at overlap threshold of 0.5 and over 75% recall at the more challenging overlap of 0.7. Our approach runs in 0.25 seconds and we additionally demonstrate a near real-time variant with only minor loss in accuracy.
2,892 citations
Posted Content•
TL;DR: This paper forms the problem of predicting local edge masks in a structured learning framework applied to random decision forests and develops a novel approach to learning decision trees robustly maps the structured labels to a discrete space on which standard information gain measures may be evaluated.
Abstract: Edge detection is a critical component of many vision systems, including object detectors and image segmentation algorithms. Patches of edges exhibit well-known forms of local structure, such as straight lines or T-junctions. In this paper we take advantage of the structure present in local image patches to learn both an accurate and computationally efficient edge detector. We formulate the problem of predicting local edge masks in a structured learning framework applied to random decision forests. Our novel approach to learning decision trees robustly maps the structured labels to a discrete space on which standard information gain measures may be evaluated. The result is an approach that obtains realtime performance that is orders of magnitude faster than many competing state-of-the-art approaches, while also achieving state-of-the-art edge detection results on the BSDS500 Segmentation dataset and NYU Depth dataset. Finally, we show the potential of our approach as a general purpose edge detector by showing our learned edge models generalize well across datasets.
670 citations
29 Sep 2014
TL;DR: This work presents a novel approach for detecting objects and estimating their 3D pose in single images of cluttered scenes using a deformable parts-based model and demonstrates successful grasps using the detection and pose estimate with a PR2 robot.
Abstract: We present a novel approach for detecting objects and estimating their 3D pose in single images of cluttered scenes. Objects are given in terms of 3D models without accompanying texture cues. A deformable parts-based model is trained on clusters of silhouettes of similar poses and produces hypotheses about possible object locations at test time. Objects are simultaneously segmented and verified inside each hypothesis bounding region by selecting the set of superpixels whose collective shape matches the model silhouette. A final iteration on the 6-DOF object pose minimizes the distance between the selected image contours and the actual projection of the 3D model. We demonstrate successful grasps using our detection and pose estimate with a PR2 robot. Extensive evaluation with a novel ground truth dataset shows the considerable benefit of using shape-driven cues for detecting objects in heavily cluttered scenes.
222 citations
Posted Content•
TL;DR: In this paper, the authors propose a new architecture for difficult image processing operations, such as natural edge detection or thin object segmentation, based on a simple combination of convolutional neural networks with the nearest neighbor search.
Abstract: We propose a new architecture for difficult image processing operations, such as natural edge detection or thin object segmentation. The architecture is based on a simple combination of convolutional neural networks with the nearest neighbor search.
We focus our attention on the situations when the desired image transformation is too hard for a neural network to learn explicitly. We show that in such situations, the use of the nearest neighbor search on top of the network output allows to improve the results considerably and to account for the underfitting effect during the neural network training. The approach is validated on three challenging benchmarks, where the performance of the proposed architecture matches or exceeds the state-of-the-art.
202 citations
28 Aug 2014
TL;DR: An improved Canny edge detection algorithm based on adaptive smooth filtering is proposed in this article, according to the revulsion characteristic of image pixels gray scale, this algorithm adaptively changes the coefficients of the filter.
Abstract: This paper introduces the fundamental theory of Canny operator and carries on its analysis and evaluation.On this foundation,an improved Canny edge detection algorithm based on adaptive smooth filtering is proposed.According to the revulsion characteristic of image pixels gray scale,this algorithm adaptively changes the coefficients of the filter.The results of the experiment pictures indicate that the improved algorithm has better accuracy and precision in the edge detection.
189 citations
01 Nov 2014
TL;DR: A new architecture for difficult image processing operations, such as natural edge detection or thin object segmentation, is proposed based on a simple combination of convolutional neural networks with the nearest neighbor search.
Abstract: We propose a new architecture for difficult image processing operations, such as natural edge detection or thin object segmentation. The architecture is based on a simple combination of convolutional neural networks with the nearest neighbor search.
178 citations
TL;DR: In this paper, two different types of cameras are used to monitor the response of a bridge to a train pass-by, and the acquired images are analyzed using three different image processing techniques (Pattern Matching, Edge Detection and Digital Image Correlation) and the results are compared with a reference measurement, obtained by a laser interferometer providing single point measurements.
Abstract: Bridge static and dynamic vibration monitoring is a key activity for both safety and maintenance purposes. The development of vision-based systems allows to use this type of devices for remote estimation of a bridge vibration, simplifying the measuring system installation. The uncertainty of this type of measurements is strongly related to the experimental conditions (mainly the pixel-to-millimeters conversion, the target texture, the camera characteristics and the image processing technique). In this paper two different types of cameras are used to monitor the response of a bridge to a train pass-by. The acquired images are analyzed using three different image processing techniques (Pattern Matching, Edge Detection and Digital Image Correlation) and the results are compared with a reference measurement, obtained by a laser interferometer providing single point measurements. Tests with different zoom levels are shown and the corresponding uncertainty values are estimated. As the zoom level decreases it is possible not only to measure the displacement of one point of the bridge, but also to grab images from a wide structure portion in order to recover displacements of a large number of points in the field of view. The extreme final solution would be having wide area measurements with no targets, to make measurements really easy, with clear advantages, but also with some drawbacks in terms of uncertainty to be fully comprehended.
165 citations
TL;DR: A distributed Canny edge detection algorithm that adaptively computes the edge detection thresholds based on the block type and the local distribution of the gradients in the image block to have a significantly reduced latency and can be easily integrated with other block-based image codecs.
Abstract: The Canny edge detector is one of the most widely used edge detection algorithms due to its superior performance. Unfortunately, not only is it computationally more intensive as compared with other edge detection algorithms, but it also has a higher latency because it is based on frame-level statistics. In this paper, we propose a mechanism to implement the Canny algorithm at the block level without any loss in edge detection performance compared with the original frame-level Canny algorithm. Directly applying the original Canny algorithm at the block-level leads to excessive edges in smooth regions and to loss of significant edges in high-detailed regions since the original Canny computes the high and low thresholds based on the frame-level statistics. To solve this problem, we present a distributed Canny edge detection algorithm that adaptively computes the edge detection thresholds based on the block type and the local distribution of the gradients in the image block. In addition, the new algorithm uses a nonuniform gradient magnitude histogram to compute block-based hysteresis thresholds. The resulting block-based algorithm has a significantly reduced latency and can be easily integrated with other block-based image codecs. It is capable of supporting fast edge detection of images and videos with high resolutions, including full-HD since the latency is now a function of the block size instead of the frame size. In addition, quantitative conformance evaluations and subjective tests show that the edge detection performance of the proposed algorithm is better than the original frame-based algorithm, especially when noise is present in the images. Finally, this algorithm is implemented using a 32 computing engine architecture and is synthesized on the Xilinx Virtex-5 FPGA. The synthesized architecture takes only 0.721 ms (including the SRAM read/write time and the computation time) to detect edges of \(512\times 512\) images in the USC SIPI database when clocked at 100 MHz and is faster than existing FPGA and GPU implementations.
149 citations
TL;DR: Experimental results have shown that the proposed technique performs better or at least at par with the state-of-the-art steganography techniques but provides higher embedding capacity.
Abstract: This paper proposes a novel steganography technique, where edges in the cover image have been used to embed messages. Amount of data to be embedded plays an important role on the selection of edges, i.e., the more the amount of data to be embedded, larger the use of weaker edges for embedding. Experimental results have shown that the proposed technique performs better or at least at par with the state-of-the-art steganography techniques but provides higher embedding capacity.
93 citations
TL;DR: This paper compares each of these operators by the manner of checking Peak signal to Noise Ratio (PSNR) and Mean Squared Error (MSE) of resultant image and finds Canny operator found as the best among others in edge detection accuracy.
Abstract: Edge detection is the vital task in digital image processing. It makes the image segmentation and pattern recognition more comfort. It also helps for object detection. There are many edge detectors available for pre-processing in computer vision. But, Canny, Sobel, Laplacian of Gaussian (LoG), Robert's and Prewitt are most applied algorithms. This paper compares each of these operators by the manner of checking Peak signal to Noise Ratio (PSNR) and Mean Squared Error (MSE) of resultant image. It evaluates the performance of each algorithm with Matlab and Java. The set of four universally standardized test images are used for the experimentation. The PSNR and MSE results are numeric values, based on that, performance of algorithms identified. The time required for each algorithm to detect edges is also documented. After the Experimentation, Canny operator found as the best among others in edge detection accuracy. Index Terms—Canny operator, Edge Detectors, Laplacian of Gaussian, MSE, PSNR, Sobel operator.
86 citations
Patent•
20 Aug 2014
TL;DR: In this article, an apparatus that generates an RGB pattern data from an image pickup signal by an image element having an RGBW pattern and a method is presented, where an edge detection unit analyzes an output signal of the image pickup signals of the RGBW patterns to obtain edge information corresponding to the respective pixels, and a texture detection unit generates texture information.
Abstract: To provide an apparatus that generates an RGB pattern data from an image pickup signal by an image pickup element having an RGBW pattern and a method. An edge detection unit analyzes an output signal of the image pickup signal of the RGBW pattern to obtain edge information corresponding to the respective pixels, and a texture detection unit generates texture information. Furthermore, a parameter calculation unit executes an interpolation processing in which an applied pixel position is changed in accordance with an edge direction of a transform target pixel to generate parameters equivalent to an interpolation pixel value. In a blend processing unit, the parameters generated by the parameter calculation unit, the edge information, and the texture information are input, in accordance with the edge information and the texture information corresponding to the transform pixel, a blend ratio of the parameters calculated by the parameter calculation unit is changed, the blend processing is executed, and a transform pixel value is decided.
01 Jan 2014
TL;DR: 6 morphological operations which are implemented in the matlab program, including erosion, dilation, opening, closing, boundary extraction and region filling are covered.
Abstract: Image processing including noise suppression, feature extraction, edge detection, image segmentation, shape recognition, texture analysis, image restoration and reconstruction, image compression etc uses mathematical morphology which is a method of nonlinear filters. It is modulated from traditional morphology to order morphology, soft mathematical morphology and fuzzy soft mathematical morphology. This paper is covers 6 morphological operations which are implemented in the matlab program, including erosion, dilation, opening, closing, boundary extraction and region filling.
TL;DR: A novel Color Global and Local Oriented Edge Magnitude Pattern (Color Global LOEMP) is proposed that is able to effectively combine color, global spatial structure, global direction structure, and local shape information and balance the two concerns of distinctiveness and robustness.
Abstract: Most of the existing traffic sign recognition (TSR) systems make use of the inner region of the signs or the local features such as Haar, histograms of oriented gradients (HOG), and scale-invariant feature transform for recognition, whereas these features are still limited to deal with the rotation, illumination, and scale variations situations. A good feature of a traffic sign is desired to be discriminative and robust. In this paper, a novel Color Global and Local Oriented Edge Magnitude Pattern (Color Global LOEMP) is proposed. The Color Global LOEMP is a framework that is able to effectively combine color, global spatial structure, global direction structure, and local shape information and balance the two concerns of distinctiveness and robustness. The contributions of this paper are as follows: 1) color angular patterns are proposed to provide the color distinguishing information; 2) a context frame is established to provide global spatial information, due to the fact that the context frame is established by the shape of the traffic sign, thus allowing the cells to be aligned well with the inside part of the traffic sign even when rotation and scale variations occur; and 3) a LOEMP is proposed to represent each cell. In each cell, the distribution of the orientation patterns is described by the HOG feature, and then, each direction of HOG is represented in detail by the occurrence of local binary pattern histogram in this direction. Experiments are performed to validate the effectiveness of the proposed approach with TSR systems, and the experimental results are satisfying, even for images containing traffic signs that have been rotated, damaged, altered in color, or undergone affine transformations or images that were photographed under different weather or illumination conditions.
Patent•
18 Mar 2014TL;DR: In this paper, a combination of known techniques from statistics, signal processing and machine learning can be used to identify outliers on unsupervised data, and to capture anomalies like edge detection, spike detection, and pattern error anomalies.
Abstract: Examining time series sequences representing performance counters from executing programs can provide significant clues about potential malfunctions, busy periods in terms of traffic on networks, intensive processing cycles and so on. An unsupervised anomaly detector can detect anomalies for any time series. A combination of known techniques from statistics, signal processing and machine learning can be used to identify outliers on unsupervised data, and to capture anomalies like edge detection, spike detection, and pattern error anomalies. Boolean and probabilistic results concerning whether an anomaly was detected can be provided.
TL;DR: A modified approach accounting for the fact that cell motility occurs over a much shorter time scale than proliferation is presented, which shows that this approach is fast, inexpensive, non-destructive and avoids the need for cell labelling and cell counting.
Abstract: Moving cell fronts are an essential feature of wound healing, development and disease. The rate at which a cell front moves is driven, in part, by the cell motility, quantified in terms of the cell...
TL;DR: The proposed scheme outperforms other previously developed methods in detecting salient regions of the static and dynamic scenes and can be easily extended to various applications, such as image retargeting, object segmentation, and video surveillance.
Abstract: Saliency detection has been extensively studied due to its promising contributions for various computer vision applications. However, most existing methods are easily biased toward edges or corners, which are statistically significant, but not necessarily relevant. Moreover, they often fail to find salient regions in complex scenes due to ambiguities between salient regions and highly textured backgrounds. In this paper, we present a novel unified framework for spatiotemporal saliency detection based on textural contrast. Our method is simple and robust, yet biologically plausible; thus, it can be easily extended to various applications, such as image retargeting, object segmentation, and video surveillance. Based on various datasets, we conduct comparative evaluations of 12 representative saliency detection models presented in the literature, and the results show that the proposed scheme outperforms other previously developed methods in detecting salient regions of the static and dynamic scenes.
TL;DR: A new and comprehensive framework for granular object recognition is established, which better incorporates intensity gradient information into the geometrical watershed framework and revise the marker-finding procedure to incorporate a clustering step in the marker finding procedure.
01 Dec 2014
TL;DR: This paper proposed an automatic method for accurate edge detection of concrete cracks from real 2D images of concrete surfaces containing noisy and unintended objects and demonstrated the excellent performance of the proposed method.
Abstract: The automatic edge detection of cracks on concrete structures plays an important role in the damage assessment process for cracked structures. In this paper, we proposed an automatic method for accurate edge detection of concrete cracks from real 2D images of concrete surfaces containing noisy and unintended objects. In the 2D image of a damaged concrete surface, cracks are usually observed as tree-like topology dark objects of which the branches are line-like and have local symmetry across their center axes. We utilize these two geometric properties of cracks to detect crack edges and discriminate them with edges of other unintended objects. The novel automatic crack edge detection is composed of two sequential stages. In the first stage, cracks are enhanced by a novel phase symmetry-based crack enhancement filter (PSCEF) based on their symmetric and line-like properties while non-crack objects are removed. Estimated crack center-lines are then obtained by thresholding the filtered images and applying morphological thinning algorithm to the binary image. In the second stage, the estimated center lines of the detected cracks are fitted by cubic splines and the pixel intensity profiles in the directions perpendicular to the splines are used to determine the edge points. The edge points are linked together to form the desired continuous crack edges. Various experiments of real concrete crack images are used to demonstrate the excellent performance of the proposed method.
TL;DR: Experiments show that the proposed approach can effectively reduce the influence of illumination and noise, contributing a robust weld line detection and tracking system.
Abstract: Unlike weld seam detection in a welding process, weld line localization for inspection is usually performed outdoors and challenged by noise and variation of illumination intensity. In this paper, we propose a weld line localization approach for mobile platform via a cross structured light (CSL) device and spatial-temporal cascaded hidden Markov models (HMMs). A CSL device is designed to project cross red laser stripes on weldment surfaces and capture the weld convexity in video sequences. Stripe edge images are extracted and then a spatial HMM is designed to detect the regions of interest (ROIs) in the video frames. Detected ROIs in successive video frames are fed to the proposed temporal HMM as observations to track the weld lines. In this way, we incorporate both the spatial characteristics of laser stripes and the continuity of the weld lines in an optimal framework. Experiments show that the proposed approach can effectively reduce the influence of illumination and noise, contributing a robust weld line detection and tracking system.
TL;DR: This paper presents a biometric technique for identification of a person using the iris image that is first segmented from the acquired image of an eye using an edge detection algorithm, and exhibits an accuracy of 98.5%.
Abstract: This paper presents a biometric technique for identification of a person using the iris image. The iris is first segmented from the acquired image of an eye using an edge detection algorithm. The disk shaped area of the iris is transformed into a rectangular form. Described moments are extracted from the grayscale image which yields a feature vector containing scale, rotation, and translation invariant moments. Images are clustered using the k-means algorithm and centroids for each cluster are computed. An arbitrary image is assumed to belong to the cluster whose centroid is the nearest to the feature vector in terms of Euclidean distance computed. The described model exhibits an accuracy of 98.5%.
TL;DR: A freely available Matlab based toolbox TecLines (Tectonic Lineament Analysis) for locating and quantifying lineament patterns using satellite data and digital elevation models and the application of the tensor-voting framework to improve position and length accuracies of the detected lineaments is introduced.
Abstract: Geological structures, such as faults and fractures, appear as image discontinuities or lineaments in remote sensing data. Geologic lineament mapping is a very important issue in geo-engineering, especially for construction site selection, seismic, and risk assessment, mineral exploration and hydrogeological research. Classical methods of lineaments extraction are based on semi-automated (or visual) interpretation of optical data and digital elevation models. We developed a freely available Matlab based toolbox TecLines (Tectonic Lineament Analysis) for locating and quantifying lineament patterns using satellite data and digital elevation models. TecLines consists of a set of functions including frequency filtering, spatial filtering, tensor voting, Hough transformation, and polynomial fitting. Due to differences in the mathematical background of the edge detection and edge linking procedure as well as the breadth of the methods, we introduce the approach in two-parts. In this first study, we present the steps that lead to edge detection. We introduce the data pre-processing using selected filters in spatial and frequency domains. We then describe the application of the tensor-voting framework to improve position and length accuracies of the detected lineaments. We demonstrate the robustness of the approach in a complex area in the northeast of Afghanistan using a panchromatic QUICKBIRD-2 image with 1-meter resolution. Finally, we compare the results of TecLines with manual lineament extraction, and other lineament extraction algorithms, as well as a published fault map of the study area.
03 Apr 2014
TL;DR: A design of a Sobel edge detection algorithm to find edge pixels in gray scale image using Xilinx ISE Design Suite-14 software platforms and VHDL language is presented.
Abstract: Real-time image processing applications requires processing on large data of pixels in a given timing constraints. Reconfigurable device like FPGAs have emerged as promising solutions for reducing execution times by deploying parallelism techniques in image processing algorithms. Implementation of highly parallel system architecture, parallel access of large internal memory banks and optimization of processing element for applications makes FPGA an ideal device for image processing system. Edge detection is basic tool used in many image processing applications for extracting information from image. Sobel edge detection is gradient based edge detection method used to find edge pixels in image. This paper presents a design of a Sobel edge detection algorithm to find edge pixels in gray scale image. Xilinx ISE Design Suite-14 software platforms is used to design a algorithm using VHDL language. MATLAB software platform is used for obtaining pixel data matrix from gray scale image and vice versa. Xilinx FPGAs of family Vertex-5 are more suitable for image processing work than Spartan-3 and Spartan-6.
03 Jun 2014
TL;DR: A robust road lane marker detection algorithm to detect the left and right lane markers based on real-time video processing and Canny edge detection and Hough Transform is presented.
Abstract: Lane detection plays an important role in intelligent vehicle systems. Therefore, this paper presents a robust road lane marker detection algorithm to detect the left and right lane markers. The algorithm consists of optimization of Canny edge detection and Hough Transform. The system captures images from a front viewing vision sensor placed facing the road behind the windscreen as input. Then a series of image processing is applied to generate the road model. Canny edge detection performs features recognition then followed by Hough Transform lane generation. The algorithm detects visible left and right lane markers on the road based on real-time video processing.
TL;DR: This study investigates the use of colour vector angle in image hashing and proposes a robust hashing algorithm combining colour vector angles with discrete wavelet transform (DWT), which is robust against normal digital operations.
Abstract: Colour vector angle has been widely used in edge detection and image retrieval, but its investigation in image hashing is still limited. In this study, the authors investigate the use of colour vector angle in image hashing and propose a robust hashing algorithm combining colour vector angles with discrete wavelet transform (DWT). Specifically, the input image is firstly resized to a normalised size by bi-cubic interpolation and blurred by a Gaussian low-pass filter. Colour vector angles are then calculated and divided into non-overlapping blocks. Next, block means of colour vector angles are extracted to form a feature matrix, which is further compressed by DWT. Image hash is finally formed by those DWT coefficients in the LL sub-band. Experiments show that the proposed hashing is robust against normal digital operations, such as JPEG compression, watermarking embedding and rotation within 5°. Receiver operating characteristics curve comparisons are conducted and the results show that the proposed hashing is better than some well-known algorithms.
20 Mar 2014
TL;DR: In this article, the authors adopt an edge detection method, establish an edge map, and adaptively select the candidate modes using the edge map for a block, and determine the number of candidate modes through trade-off between computational complexity and coding efficiency.
Abstract: High efficiency video coding (HEVC) appears due to the demand on high compression video coding beyond H.264/AVC in ultra-high definition (UHD) videos, and it brings high computational complexity with a variety of state of the art coding tools. As for intra prediction, HEVC has 35 prediction modes while H.264/AVC has 9 intra modes. To exploit the spatial correlation, we adopt an edge detection method, establish an edge map, and adaptively select the candidate modes using the edge map for a block. The number of the candidate modes is determined through trade-off between computational complexity and coding efficiency. Besides, the range of coding unit sizes is determined using the uniqueness of the edge directions for the given image block. The proposed scheme reduced the encoding time by 56.8% at the cost of 2.5% BD-BR increase on average compared to Full modes at the HEVC reference software (HM 10.0 [1]).
TL;DR: The proposed algorithm can not only effectively suppress speckle noise to improve the PSNR of SAR image, but also significantly improves the visual effect of SAR images, especially in enhancing the image’s texture.
Abstract: As SAR has been widely used nearly in every field, how to improve SAR's image in both quality and visual effect has become necessary. Before what we really process the SAR image like image segmentation, edge detection, target detection or other processing, we must suppress the speckle noise in the image firstly. By analyzing the sorts and origins of noises, we present a new de-noising method of SAR image in the Shearlet domain based on sparse representation and Bayesian theory. Firstly, we apply the Shearlet transform to the noised SAR image. Secondly, we construct a new de-noising model via sparse representation and then use iterative algorithm based on Bayesian theory to solve it. Lastly, we can obtain the clean SAR image from the de-nosing Shearlet coefficients. The experimental results show that the proposed algorithm can not only effectively suppress speckle noise to improve the PSNR of SAR image, but also significantly improves the visual effect of SAR image, especially in enhancing the image's texture.
TL;DR: An automatic algorithm for segmenting retinal layers based on dual-gradient and spatial correlation smoothness constraint is developed and demonstrated that the proposed method can estimate six layer boundaries accurately.
TL;DR: In this paper, the authors proposed a novel method to detect edge on clear or noisy images based on neutrosophic set and a new directional α-mean operation is defined, which performs well on images without noise or with different levels of noise.
TL;DR: This paper proposes a new approach to label-equivalence-based two-scan connected-component labeling that uses two strategies to reduce repeated checking-pixel work for labeling and was more efficient than all conventional labeling algorithms.
Abstract: This paper proposes a new approach to label-equivalence-based two-scan connected-component labeling. We use two strategies to reduce repeated checking-pixel work for labeling. The first is that instead of scanning image lines one by one and processing pixels one by one as in most conventional two-scan labeling algorithms, we scan image lines alternate lines, and process pixels two by two. The second is that by considering the transition of the configuration of pixels in the mask, we utilize the information detected in processing the last two pixels as much as possible for processing the current two pixels. With our method, any pixel checked in the mask when processing the current two pixels will not be checked again when the next two pixels are processed; thus, the efficiency of labeling can be improved. Experimental results demonstrated that our method was more efficient than all conventional labeling algorithms.
TL;DR: An accurate iris localization and high recognition performance approach for noisy iris images is presented and the thorough experimental results on the challenging iris image database CASIA-Iris-Thousand achieve an EER of 1.8272 %, which outperforms the state-of-the-art methods.
Abstract: Iris recognition plays an important role in biometrics. Until now, many scholars have made different efforts in this field. However, the recognition performances of most proposed methods degrade dramatically when the image contains some noise, which inevitably occurs during image acquisition such as reflection spots, inconsistent illumination, eyelid, eyelash, hair, etc. In this paper, an accurate iris localization and high recognition performance approach for noisy iris images is presented. After filling the reflection spots using the inpainting method which is based on Navier-Stokes (NS) equations, the Probable boundary (Pb) edge detection operator is used to detect pupil edge initially, which can eliminate the interference of inconsistent illumination, eyelid, eyelash and hair. Besides, the accurate circle parameters are obtained in delicately to reduce the input space of Hough transforms. The iris feature code is constructed based on 1D Log-Gabor filter. Our thorough experimental results on the challenging iris image database CASIA-Iris-Thousand achieve an EER of 1.8272 %, which outperforms the state-of-the-art methods.