PixISegNet: pixel-level iris segmentation network using convolutional encoder–decoder with stacked hourglass bottleneck

In this paper, we present a new iris ROI segmentation algorithm using a deep convolutional neural network (NN) to achieve the state-of-the-art segmentation performance on well-known iris image data sets. The authors’ model surpasses the performance of state-of-the-art Iris DenseNet framework by appl...

Celý popis

Uložené v:

Podrobná bibliografia
Vydané v:	IET biometrics Ročník 9; číslo 1; s. 11 - 24
Hlavní autori:	Jha, Ranjeet Ranjan, Jaswal, Gaurav, Gupta, Divij, Saini, Shreshth, Nigam, Aditya
Médium:	Journal Article
Jazyk:	English
Vydavateľské údaje:	Stevenage The Institution of Engineering and Technology 01.01.2020 John Wiley & Sons, Inc
Predmet:	Algorithms Artificial neural networks Aspect ratio biometric segmentation research Biometrics Blurring CASIA V3.0 interval data sets Coders content loss optimisation convolutional neural nets cross‐entropy loss Datasets decoding deep convolutional neural network deep convolutional NN Deep learning encoder–decoder Entropy entropy codes Feature extraction hyper‐parameterisation IIT‐D interval data sets image classification image coding Image contrast image matching Image processing image restoration Image segmentation image segmentation performance Iris iris image data sets iris recognition iris ROI image segmentation algorithm Iris‐DenseNet framework Localization Methods multiscale‐multiorientation training Neural networks nonregular reflections Parameterization Pixels pixel‐level iris segmentation network pixel‐to‐pixel classification loss Pix‐SegNet Research Article salient iris features Segmentation siamese matching network stacked hourglass bottleneck stacked hourglass network TOTA strategy train‐once‐test‐all strategy UBIRIS‐V2 interval data sets Wavelet transforms train-once-test-all strategy iris image data sets image classification iris ROI image segmentation algorithm multiscale-multiorientation training UBIRIS-V2 interval data sets deep convolutional NN siamese matching network content loss optimisation image restoration image segmentation feature extraction Pix-SegNet IIT-D interval data sets Iris-DenseNet framework deep convolutional neural network CASIA V3.0 interval data sets image segmentation performance entropy codes image matching decoding cross-entropy loss stacked hourglass bottleneck nonregular reflections stacked hourglass network TOTA strategy encoder–decoder hyper-parameterisation pixel-level iris segmentation network iris recognition convolutional neural nets pixel-to-pixel classification loss biometric segmentation research image coding salient iris features
ISSN:	2047-4938, 2047-4946, 2047-4946
On-line prístup:	Získať plný text
Tagy:	Pridať tag Žiadne tagy, Buďte prvý, kto otaguje tento záznam!

Popis
Shrnutí:	In this paper, we present a new iris ROI segmentation algorithm using a deep convolutional neural network (NN) to achieve the state-of-the-art segmentation performance on well-known iris image data sets. The authors’ model surpasses the performance of state-of-the-art Iris DenseNet framework by applying several strategies, including multi-scale/ multi-orientation training, model training from scratch, and proper hyper-parameterisation of crucial parameters. The proposed PixISegNet consists of an autoencoder which primarily uses long and short skip connections and a stacked hourglass network between encoder and decoder. There is a continuous scale up–down in stacked hourglass networks, which helps in extracting features at multiple scales and robustly segments the iris even in an occluded environment. Furthermore, cross-entropy loss and content loss optimise the proposed model. The content loss considers the high-level features, thus operating at a different scale of abstraction, which compliments the cross-entropy loss, which considers pixel-to-pixel classification loss. Additionally, they have checked the robustness of the proposed network by rotating images to certain degrees with a change in the aspect ratio along with blurring and a change in contrast. Experimental results on the various iris characteristics demonstrate the superiority of the proposed method over state-of-the-art iris segmentation methods considered in this study. In order to demonstrate the network generalisation, they deploy a very stringent TOTA (i.e. train-once-test-all) strategy. Their proposed method achieves $E_1$E1 scores of 0.00672, 0.00916 and 0.00117 on UBIRIS-V2, IIT-D and CASIA V3.0 Interval data sets, respectively. Moreover, such a deep convolutional NN for segmentation when included in an end-to-end iris recognition system with a siamese based matching network will augment the performance of the siamese network.
Bibliografia:	ObjectType-Article-1 SourceType-Scholarly Journals-1 ObjectType-Feature-2 content type line 14
ISSN:	2047-4938 2047-4946 2047-4946
DOI:	10.1049/iet-bmt.2019.0025