Iris Segmentation Research Articles

Hybrid Iris Segmentation Method Based on CNN and Principal Curvatures

In this article the new hybrid iris image segmentation method based on convolutional neural networks and mathematical methods is proposed. Iris boundaries are found using modified Daugman’s method. Two UNet-based convolutional neural networks are used for iris mask detection. The first one is used to predict the preliminary iris mask including the areas of the pupil, eyelids and some eyelashes. The second neural network is applied to the enlarged image to specify thin ends of eyelashes. Then the principal curvatures method is used to combine the predicted by neural networks masks and to detect eyelashes correctly. The pro- posed segmentation algorithm is tested using images from CASIA IrisV4 Interval database. The results of the proposed method are evaluated by the Intersection over Union, Recall and Precision metrics. The average metrics values are 0.922, 0.957 and 0.962, respectively. The proposed hy- brid iris image segmentation approach demonstrates an improvement in comparison with the methods that use only neural networks.

A fast specular reflection removal based on pixels properties method

Iris recognition has been around for many years due to an extensive research on the uniqueness of human iris. It is well known that the iris is not similar to each other which means every human in the planet has their own iris pattern and cannot be shared. One of the main issues in iris recognition is iris segmentation. One element that can reduce the accuracy of iris segmentation is the presence of specular reflection. Another issue is the speed of specular reflection removal since the iris recognition system needs to process a lot of irises. In this paper, a specular reflection removal method was proposed to achieve a fast and accurate specular reflection removal. Some modifications were implemented on the existing pixels properties method. Based on the results, the proposed method achieved the fastest execution time, the highest segmentation accuracy and the highest SSIM compared to the other methods. This proves that the proposed method is fast and accurate to be implemented in the iris recognition systems.

I-SOCIAL-DB: A labeled database of images collected from websites and social media for Iris recognition

People upload daily a huge number of portrait face pictures on websites and social media, which can be processed using biometric systems based on the face characteristics to perform an automatic recognition of the individuals. However, the performance of face recognition approaches can be limited by negative factors as aging, occlusions, rotations, and uncontrolled expressions. Nevertheless, the constantly increasing quality and resolution of the portrait pictures uploaded on websites and social media could permit to overcome these problems and improve the robustness of biometric recognition methods by enabling the analysis of additional traits, like the iris. To point the attention of the research community to the possible use of iris-based recognition techniques for images uploaded on websites and social media, we present a public image dataset called I-SOCIAL-DB (Iris Social Database). This dataset is composed of 3,286 ocular regions, extracted from 1,643 high-resolution face images of 400 individuals, collected from public websites. For each ocular region, a human expert extracted the coordinates of the circles approximating the inner and outer iris boundaries and performed a pixelwise segmentation of the iris contours, occlusions, and reflections. This dataset is the first collection of ocular images from public websites and social media, and one of the biggest collections of manually segmented ocular images in the literature. In this paper, we also present a qualitative analysis of the samples, a set of testing protocols and figures of merit, and benchmark results achieved using publicly available iris segmentation and recognition algorithms. We hope that this initiative can give a new test tool to the biometric research community, aiming to stimulate new studies in this challenging research field.

A robust iris localization scheme for the iris recognition

Due to current security situations around the globe, iris biometric technology is highly preferred for both overt and covert applications. A typical iris biometric system includes image acquisition, iris segmentation, features extraction, and matching and recognition modules. Amongst these modules, iris segmentation plays a decisive role because it segments the valid iris part in an input eyeimage. It includes two tasks: iris localization and noise (e.g., eyelids) removal. Notably, the overall performance of an iris biometric system strongly relies on the iris localization task, because it demarcates the actual iris contours. Some contemporary iris localization schemes search over a three-dimensional (3D) space while marking iris boundaries, which is a time-consuming process if not optimized properly. Besides, some schemes also resort to the fixed and/or crude thresholding-based techniques for pupil localization. Notably, such schemes may perform poorly if image data do not maintain quality. To address these issues, this study proposes a robust iris localization scheme maintaining both speed and accuracy. It includes preprocessing the input eyeimage using an order statistic-filter and the bilinear interpolation scheme, extracting an adaptive threshold using the image’s histogram, processing binary image via the morphological operators, extracting pupil’s center and radius based on the centroid and geometry concepts, marking iris outer boundary using the Circular Hough transform (CHT) and refining coarse iris boundaries through the Fourier series. The proposed scheme exhibits relatively better experimental results compared with some contemporary iris localization schemes on the public iris databases: IITD V1.0, CASIA-Iris-Interval and MMU V1.0.

A Modified Chan-Vese Active Contour for Iris Segmentation on Non-Ideal and Noisy Iris

A Modified Chan-Vese Active Contour for Iris Segmentation on Non-Ideal and Noisy Iris

Iris Segmentation Using Feature Channel Optimization for Noisy Environments

In recent years, iris recognition has been widely used in various fields. As the first step of iris recognition, segmentation accuracy is of great significance to the final recognition. However, iris images exhibit a variety of noise in the real world, which leads to lower segmentation accuracy than the ideal case. To address this problem, this paper proposes an iris segmentation method using feature channel optimization for noisy images. The method for non-ideal environments with noise is more suitable for practical applications. We add dense blocks and dilated convolutional layers to the encoder so that the information gradient flow obtained by different layers can be reused, and the receptive field can be expanded. In the decoder, based on Jensen-Shannon (JS) divergence, we first recalculate the weight of the feature channels obtained from each layer, which enhances the useful information and suppresses the interference information in the noisy environments to boost the segmentation accuracy. The proposed architecture is validated in the CASIA v4.0 interval (CASIA) and IIT Delhi v1.0 datasets (IITD). For CASIA, the mean error rate is 0.78%, and the F-measure value is 98.21%. For IITD, the mean error rate is 0.97%, and the F-measure value is 97.87%. Experimental results show that the proposed method outperforms other state-of-art methods under noisy environments, such as Gaussian blur, Gaussian noise, and salt and pepper noise.

Investigation of iris segmentation techniques using active contours for non-cooperative iris recognition

<span>Iris image segmentation process based on graphical user interface (GUI) to accurately localize the iris structure is presented in this paper. The major challenge confronting the precision of an iris recognition model is how to determine the accuracy of the iris segmentation and localization. There are varying parameters that introduce constraints during feature extraction and these greatly affect the matching performance during iris localization. To this end, the Integro-differential operator, which involves the detection of inner and outer regions of the iris, and the circular hough transform, which is capable of detecting the circular boundary from the edge mapping were investigated, and an active contour model was evolved. In the evolved model, an emerging curve mapped with the zeros of the data set function is experimentally exploited. To demonstrate the suitability of the model for precise iris recognition, its parameters were compared against other related models. Simulation results show that the model has higher flexibility of substitution of images, and the images could be analyzed more accurately with less false rejections (FR) and false acceptance (FA) in comparison with the integro-differential operator. This implies that images could be analyzed faster using the evolved model, and easily substituted especially in situations where the need to care for numerous eye patients occur.</span>

Research on Image Preprocessing Algorithm and Deep Learning of Iris Recognition

With the development of information society, biometrics technology has been paid more and more attention. Iris recognition is considered as the most promising biometric authentication technology in the 21st century because of its uniqueness, stability and non-creativity. However, due to the high cost of iris recognition equipment and some defects of the algorithm, iris recognition cannot be applied in real life on a large scale. In this paper, a fast localization iris recognition algorithm is proposed, which combines the iris segmentation algorithm with deep learning to quickly extract the iris region for recognition. Firstly, the pupil edge was extracted by dynamic threshold analysis and contour extraction, and then iris was located by edge detection and gray calculation. Finally, features of normalized images were learned by deep learning network. Experiments show that the method can guarantee the accuracy and efficiency of iris segmentation and has a high degree of recognition and matching.

Detection of uveal melanoma using fuzzy and neural networks classifiers

The use of image processing is increasingly utilized for disease detection. In this article, an algorithm is proposed to detect uveal melanoma (UM) which is a type of intraocular cancer. The proposed method integrates algorithms related to iris segmentation and proposes a novel algorithm for the detection of UM from the approach of fuzzy logic and neural networks. The study case results show 76% correct classification in the fuzzy logic system and 96.04% for the artificial neural networks.

Iris Image Segmentation and Localization using Dynamic Reconfigurable Processor

Recognition of the Iris is among the finest techniques in the field of bio-metrics identification, because the iris has characteristics that are unique and stay the same all through the individual’s life. Iris recognition phases are namely image acquisition, segmentation of iris, localization of iris, feature extraction of iris and matching. This paper, which is an extension of the survey paper Smaran et.al[1], concentrates purely on the procedures of image capture, segmentation as well as localization of the iris. The aim of the paper is to optimize the above mentioned processes in terms of distance of capturing the image, time taken for memory and computation requirements, using the DRP (Dynamic Re-Configurable Processor) technology, uniquely developed by Renesas Electronics (www.renesas.com).

An effective iris segmentation scheme for noisy images

Iris segmentation plays a critical role in the iris biometric systems. It has two modules: iris localization and noise detection. The first module demarcates the actual iris’ inner and outer boundaries in input eyeimages. The second module detects and removes noise in the valid iris part. Researchers devised numerous iris segmentation and/or localization schemes, which are based on the histogram and thresholding, circular Hough transform (CHT), Integro-differential operator (IDO), active contour models, graph-cuts, or deep learning. It is observed that most contemporary schemes perform poorly when confronted with images containing noisy factors such as the eyebrows, eyelashes, contact lenses, non-uniform illumination, light reflections, defocus and/or eyeglasses. The performance of CHT and IDO against noise is found robust, but these operators are computationally expensive. On the other hand, the histogram and thresholding-based schemes are considered fast, but these are less robust against noise. Besides, most contemporary schemes mark iris contours with a circle approximation and offer no noise removal strategy. To address these issues, this study offers an effective iris segmentation algorithm. First, it applies an optimized coarse-to-fine scheme based on an adaptive threshold to mark iris inner boundary. Next, it detects and marks eyelashes adaptively. After that, it marks iris outer boundary via an optimized coarse-to-fine scheme. Then, it regularizes the non-circular iris’ contours using the Fourier series. Finally, eyelids and reflections are marked in the iris polar form. The proposed scheme shows better results on the CASIA-Iris-Interval V3.0, IITD V1.0, and MMU V1.0 iris databases.

Semantic Segmentation of Iris using U-Net in Deep Learning

In the field of medicine, iris segmentation has become a great field of interest from the past few years. Iris segmentation is also largely used in iris recognition systems [3] which are extensively used in security control [1][2]. Here iris segmentation is done using semantic segmentation which is based on the U-Net architecture. The typical U-net architecture contains two paths- contracting path containing convolutional and pooling layers and the expanding path consists of transposed convolutional operations. The UBIRIS dataset is trained on the traditional U-Net model with some modifications according to the size of the images present in the UBIRIS dataset. The results obtained were very close to the ground truths and accuracy obtained is also appreciable.

Development of Artificial Bee Colony Algorithm for Iris Segmentation

Iris segmentation is a crucial process in the iris recognition that it locates and captures the unique features inside the human iris image. However, iris image contains high noise rate and sometimes a genuine user is recognized as non-genuine by computer system. An algorithm from swarm-based intelligence is demanded for natural search images. Therefore, an experimental investigation was conducted to explore artificial bee colony (ABC) algorithm for iris segmentation. The ABC algorithm measures the performance of convergence in learning the authentic iris features. Based on the findings, the average fitness is always less than or equal to the best fitness, and the differences between the two goes on decreasing over time, which is until the algorithm completely converges. The parameters of ABC Algorithm are programmed with at best 100 times of calculation to collect the best iteration and best fitness. In fact, results showed that ABC uses the population for every iteration instead of a single iteration. Number of iterations directly affect the time of optimal search for ideal convergence, which meets the aim of this study. As a conclusion, if the mean and standard deviation is low in values, it is proven that the speed of searching the iris features is faster and more robust.

An Iris Recognition System Using Deep convolutional Neural Network

Machine learning rises in varied areas of computer science. A deep conventional neural network is powerful visual models of machine learning. We tend to present robustness and effective structure for the iris recognition system. The image first pass through these stages: enhancing the image quality, determine the iris and pupil center and radius for iris segmentation, converting the image from the Cartesian coordinates to the polar coordinates to reduce the time of processing. The proposed system is named IRISNet that extracting the feature and classifying them automatically without any domain knowledge. The architecture of IRISNet consists of Convolutional Neural Network layers for extract feature and Softmax layer to classify these features into N classes, for training CNN the backpropagation algorithm and Adam optimization method are used for updating the weights and learning rate, respectively. The performance of the proposed system was evaluated using IITD V1 iris database. The results obtained from the proposed system outperform supervised classification model (SVM, KNN, DT, and NB). The identification rate 97.32% and 96.43% for original and normalized images respectively. The recognition time per person is less than one second.

Iris Recognition Method for Non Ideal Images

The issue of counterfeiting is one of the most pressing threats to people’s security because of the many complex security problems. Many systems used in the field of verification and identification but the basic criterion for choosing is the degree of safety identification. Identification and verification systems use biometric characteristics of individuals. The system of iris recognition is one of the most important biometric recognition systems. The system used iris images for 100 persons belonging to the UBIRIS V1 database. Each person has 5 captures; the total number is 500 images. A hybrid iris segmentation method combined two algorithms ChanVese (CV) and grabcut to obtain high verification rate and to detect the outer boundary and localize the iris region. The system includes several steps starting with acquiring the image of the iris, enhancing the image quality, segment iris region, determining the coordinates of iris center and radius of the pupil, convert the iris coordinates of the cutting segments from Cartesian coordinates to the polar coordinates to reduce the processing time and to avoid the noise image problems due to eyelids and eyelashes, then utilize Gabor wavelets to extract the features of iris, and finally perform the matching process using Euclidian distance. The performance indicators, which include the recognition rate, false rejection rate(FRR), False Acceptance Rate(FAR), Correct identification rate(CIR), are as follows(97.5%, 39%, 0.025%, 0.0975). The proposed system has many advantages includes high processing capability of low-quality images and low storage capacity.

An Effective Model for Detection of Dysfunctionality in Heart Based on Iridology using Deep Neural Networks

In today’s world heart disease is the primary reason for deaths. WHO has anticipated that 12 million people die every year because of heart diseases. Every organ of the body is represented in the iris in a well-defined manner. The Iris is a micro-structure of the entire body. The abnormality of the heart can be detected using Iridology science. In this article, we examine the heart dysfunctionality through a chain of steps which are localization of iris, segmentation of iris, ROI extraction, histogram equalization of ROI and classification using deep convolutional neural network. The results are assessed based on various standards such as precision, recall, f-score & accuracy.

Iris segmentation using a new unsupervised neural approach

<span lang="EN-GB">A biometric system of identification and authentication provides automatic recognition of an individual based on certain unique features or characteristic possessed by an individual. Iris recognition is a biometric identification method that uses pattern recognition on the images of the iris. Owing to the unique epigenetic patterns of the iris, Iris recognition is considered as one of the most accurate methods in the field of biometric identification. One of the crucial steps in the iris recognition system is the iris segmentation because it significantly affects the accuracy of the feature extraction the iris. The segmentation algorithm proposed in this article starts with determining the regions of the eye using unsupervised neural approach, after the outline of the eye is found using the Canny edge, The Hough Transform is employed to determine the </span><span lang="EN-US">center</span><span lang="EN-GB"> and radius of the pupil and the iris.</span>

Iris segmentation using Hough Transform method and Fuzzy C-Means method

Complementary and Alternative Medicine (CAM) therapy is quite popular for chronic diseases such as diabetes, arthritis, and others. CAM has become very popular in the last period. One area of CAM is iridology, which is an alternative diagnosis that links iris patterns, color, tissue weakness, damage and other characteristics, which can obtain evidence about the patient’s systemic health. In identification based on the eye image, it requires an iris separation process. This separation process in image processing is done by segmentation. So that segmentation plays an important role in the identification of diseases based on images. In this research segmentation is done by combining the Hough Transform method and the clustering-based segmentation method, the Fuzzy C-Means method. The segmentation is done on the iris object which is sourced from UBIRIS V2 database. System evaluation is done by measuring accuracy, sensitivity, specificity and execution time. Test results on the tested iris image indicate the proposed method is able to segment the iris image properly.

Blind Quality Assessment of Iris Images Acquired in Visible Light for Biometric Recognition.

Image quality is a key issue affecting the performance of biometric systems. Ensuring the quality of iris images acquired in unconstrained imaging conditions in visible light poses many challenges to iris recognition systems. Poor-quality iris images increase the false rejection rate and decrease the performance of the systems by quality filtering. Methods that can accurately predict iris image quality can improve the efficiency of quality-control protocols in iris recognition systems. We propose a fast blind/no-reference metric for predicting iris image quality. The proposed metric is based on statistical features of the sign and the magnitude of local image intensities. The experiments, conducted with a reference iris recognition system and three datasets of iris images acquired in visible light, showed that the quality of iris images strongly affects the recognition performance and is highly correlated with the iris matching scores. Rejecting poor-quality iris images improved the performance of the iris recognition system. In addition, we analyzed the effect of iris image quality on the accuracy of the iris segmentation module in the iris recognition system.

Efficient method for segmentation of noisy and non‐circular iris images using improved particle swarm optimisation‐based MRFCM

Segmentation of the iris is a crucial stage in an automated iris-based recognition system. The performance of any biometric system primarily relies on how effectively the iris is extracted from the unwanted parts of an iris image. The process of iris segmentation is mainly affected by the noise artefacts such as eyelid/eyelashes occlusions, specular reflections, intensity inhomogeneities, and non-circularity of the iris boundary. A novel and an efficient method has been proposed in this work to segment noisy and non-circular iris boundaries. The mathematical modelling of morphological reconstruct fuzzy C-means clustering (MRFCM) has been presented. The MRFCM based on improved particle swarm optimisation has been implemented before the segmentation in the recognition framework. The resultant images are then segmented by employing geodesic active contours incorporated by a new stopping function. The effect of the proposed segmentation method on iris recognition is observed through matching score distribution. The popular and publicly available datasets such as UBIRISv1, CASIA-v3-Interval, MMU1, and Mobile Iris Challenge Evaluation databases are considered for the evaluation of the proposed method. Recognition accuracy is validated and compared with the well-existing methods.