Iris Segmentation Using Deep Neural Networks

被引：0

作者：

Sinha, Nirmitee ^{[1
]}

Joshi, Akanksha ^{[2
]}

Gangwar, Abhishek ^{[2
]}

Bhise, Archana ^{[1
]}

Saquib, Zia ^{[2
]}

机构：

[1] SVKMs NMIMS, MPSTME, Bombay, Maharashtra, India

[2] CDAC, Bombay, Maharashtra, India

来源：

2017 2ND INTERNATIONAL CONFERENCE FOR CONVERGENCE IN TECHNOLOGY (I2CT) | 2017年

关键词：

Iris Segmentation; Deep Neural Networks; Convolutional Neural Networks;

D O I：

暂无

中图分类号：

T [工业技术];

学科分类号：

08 ;

摘要：

Iris recognition is very difficult to perform as it requires an environment that is highly controlled for better image acquisition. As compared to other biometric technologies, iris recognition is prone to poor image quality. Specially, images captured from a distance introduce noises such as blur, off axis, specular reflections and occlusions. For proper recognition good quality of captured image is mandatory and hence sometimes denoising is required. The approach discussed in the paper uses deep neural network for eliminating the unwanted patches affecting the performance of iris recognition systems. The proposed model uses upsampled indices at the decoder stage which is memory efficient. The experimental analysis is performed using Ubiris V.2 database.

引用

页码：548 / 555

页数：8