CASIA-IrisV4

CASIA-IrisV4 is an extension of CASIA-IrisV3 and contains six subsets. The three subsets from CASIA-IrisV3 are CASIA-Iris-Interval, CASIA-Iris-Lamp, and CASIA-Iris-Twins respectively. The three new subsets are CASIA-Iris-Distance, CASIA-Iris-Thousand, and CASIA-Iris-Syn. CASIA-IrisV4 contains a total of 54,601 iris images from more than 1,800 genuine subjects and 1,000 virtual subjects. All iris images are 8 bit gray-level JPEG files, collected under near infrared illumination or synthesized. Some statistics and features of each subset are given in Table 1. The six data sets were collected or synthesized at different times and CASIA-Iris-Interval, CASIA-Iris-Lamp, CASIA-Iris-Distance, CASIA-Iris-Thousand may have a small inter-subset overlap in subjects. Iris images of CASIA-Iris-Interval were captured with our self-developed close-up iris camera (Fig.1). The most compelling feature of our iris camera is that we have designed a circular NIR LED array, with suitable luminous flux for iris imaging. Because of this novel design, our iris camera can capture very clear iris images (see Fig.2). CASIA-Iris-Interval is well-suited for studying the detailed texture features of iris images. CASIA-Iris-Lamp was collected using a hand-held iris sensor produced by OKI (Fig.3). A lamp was turned on/off close to the subject to introduce more intra-class variations when we collected CASIA-Iris-Lamp. Elastic deformation of iris texture (Fig.4) due to pupil expansion and contraction under different illumination conditions is one of the most common and challenging issues in iris recognition. So CASIA-Iris-Lamp is good for studying problems of non-linear iris normalization and robust iris feature representation. CASIA-Iris-Twins contains iris images of 100 pairs of twins, which were collected during Annual Twins Festival in Beijing using OKI's IRISPASS-h camera (Fig.5). Although iris is usually regarded as a kind of phenotypic biometric characteristics and even twins have their unique iris patterns, it is interesting to study the dissimilarity and similarity between iris images of twins. CASIA-Iris-Distance contains iris images captured using our self-developed long-range multi-modal biometric image acquisition and recognition system (LMBS, Fig.6). The advanced biometric sensor can recognize users from 3 meters away by actively searching iris, face or palmprint patterns in the visual field via an intelligent multi-camera imaging system. The LMBS is human-oriented by fusing computer vision, human computer interaction and multi-camera coordination technologies and improves greatly the usability of current biometric systems. The iris images of CASIA-Iris-Distance were captured by a high resolution camera so both dual-eye iris and face patterns are included in the image region of interest (Fig. 7).And detailed facial features such as skin pattern are also visible for multi-modal biometric information fusion. CASIA-Iris-Thousand contains 20,000 iris images from 1,000 subjects, which were collected using IKEMB-100 camera (Fig. 8) produced by IrisKing. IKEMB-100 is a dual-eye iris camera with friendly visual feedback, realizing the effect of ¡°What You See Is What You Get¡±. The bounding boxes shown in the frontal LCD help users adjust their pose for high-quality iris image acquisition. The main sources of intra-class variations in CASIA-Iris-Thousand are eyeglasses and specular reflections. Since CASIA-Iris-Thousand is the first publicly available iris dataset with one thousand subjects, it is well-suited for studying the uniqueness of iris features and develop novel iris classification and indexing methods. CASIA-Iris-Syn contains 10,000 synthesized iris images of 1,000 classes. The iris textures of these images are synthesized automatically from a subset of CASIA-IrisV1 with the approach described in [1] (Fig. 10). Then the iris ring regions were embedded into the real iris images, which makes the artificial iris images more realistic. The intra-class variations introduced into the synthesized iris dataset include deformation, blurring, and rotation, which raise a challenge problem for iris feature representation and matching. We have demonstrated in [1] that the synthesized iris images are visually realistic and most subjects can not distinguish genuine and artificial iris images. More importantly, the performance results tested on the synthesized iris image database have similar statistical characteristics to genuine iris database. So users of CASIA-IrisV4 are encouraged to use CASIA-Iris-Syn for iris recognition research and any suggestions are welcome. If CASIA-Iris-Syn proves to be successful for most researchers of iris recognition, we will provide more and more synthesized iris images in the future.

CASIA-IrisV4是CASIA-IrisV3的扩展版本，共包含六个子集。其中源自CASIA-IrisV3的三个子集分别为CASIA-Iris-Interval、CASIA-Iris-Lamp与CASIA-Iris-Twins，新增的三个子集为CASIA-Iris-Distance、CASIA-Iris-Thousand及CASIA-Iris-Syn。
该数据集总计包含来自1800余名真实受试者与1000名虚拟受试者的54601幅虹膜图像。所有虹膜图像均为8位灰度JPEG（JPEG）文件，采集于近红外光照条件下或通过合成生成。各子集的部分统计信息与特性详见表1。六个数据集于不同时段采集或合成，且CASIA-Iris-Interval、CASIA-Iris-Lamp、CASIA-Iris-Distance、CASIA-Iris-Thousand在受试者层面存在少量子集间重叠。
CASIA-Iris-Interval的虹膜图像由我们自主研发的近距虹膜相机采集（图1）。该相机的核心优势在于设计了环形近红外发光二极管（NIR LED）阵列，其光通量适配虹膜成像需求。凭借这一创新设计，本相机可采集清晰度极高的虹膜图像（见图2）。CASIA-Iris-Interval非常适合用于研究虹膜图像的精细化纹理特征。
CASIA-Iris-Lamp采用OKI公司生产的手持式虹膜传感器采集（图3）。在该数据集的采集过程中，我们通过在受试者附近开启或关闭光源，引入了更多类内变异。由于不同光照条件下瞳孔扩张与收缩会导致虹膜纹理发生弹性形变（图4），这是虹膜识别领域最常见且极具挑战性的问题之一。因此CASIA-Iris-Lamp适用于研究非线性虹膜归一化与鲁棒性虹膜特征表示相关问题。
CASIA-Iris-Twins包含100对双胞胎的虹膜图像，采集自北京双胞胎节，使用的是OKI公司的IRISPASS-h相机（图5）。尽管虹膜通常被视为一种表型生物特征，即便双胞胎也拥有独一无二的虹膜纹理，但研究双胞胎虹膜图像之间的差异性与相似性仍具有重要研究价值。
CASIA-Iris-Distance的虹膜图像由我们自主研发的远程多模态生物特征图像采集与识别系统（LMBS，图6）采集。这套先进的生物传感器可通过智能多相机成像系统主动搜索视野内的虹膜、人脸或掌纹特征，实现3米外的用户身份识别。LMBS融合了计算机视觉、人机交互与多相机协同技术，以用户为中心，大幅提升了现有生物识别系统的易用性。该数据集的虹膜图像由高分辨率相机采集，感兴趣区域同时包含双眼虹膜与面部特征（图7），诸如皮肤纹理等精细化面部特征也清晰可见，可用于多模态生物特征信息融合研究。
CASIA-Iris-Thousand包含来自1000名受试者的20000幅虹膜图像，由IrisKing公司生产的IKEMB-100相机（图8）采集。IKEMB-100是一款具备友好视觉反馈功能的双眼虹膜相机，实现了“所见即所得”的采集效果。前置液晶显示屏上的边框框选区域可帮助用户调整姿态，以获取高质量的虹膜图像。该数据集类内变异的主要来源为眼镜与镜面反射。作为首个公开可用的包含1000名受试者的虹膜数据集，CASIA-Iris-Thousand非常适合用于研究虹膜特征的唯一性，以及开发新型虹膜分类与索引方法。
CASIA-Iris-Syn包含1000个类别的10000幅合成虹膜图像。这些图像的虹膜纹理基于CASIA-IrisV1的子集，通过文献[1]中提及的方法自动合成（图10）。随后将虹膜环形区域嵌入真实虹膜图像中，使人工合成的虹膜图像更具真实感。该合成数据集引入的类内变异包括形变、模糊与旋转，为虹膜特征表示与匹配带来了挑战。我们在文献[1]中已证明，合成虹膜图像在视觉上极具真实性，多数受试者无法区分真实与合成虹膜图像。更重要的是，在该合成数据集上测试得到的性能结果，与真实虹膜数据库的统计特征高度相似。因此我们鼓励CASIA-IrisV4的使用者采用CASIA-Iris-Syn开展虹膜识别研究，并欢迎提出相关建议。若该数据集能为多数虹膜识别研究者带来助力，我们未来将提供更多的合成虹膜图像。

搜集汇总

背景与挑战

背景概述

CASIA-IrisV4是一个虹膜图像数据集，包含六个子集，总计54,601张图像，适用于虹膜识别研究。数据集包含真实和虚拟受试者的虹膜图像，采集于不同条件下，支持多种虹膜识别相关研究。