Efficient and Accurate Iris Detection and Segmentation Based on Multi-scale Optimized Mask R-CNN

摘要

Iris segmentation plays an important role in iris recognition. However, traditional iris segmentation performance decreases dramatically on non-constrained conditions, which stops iris recognition system from being widely deployed. In this paper, an efficient and accurate iris detection and segmentation method based on multi-scale optimized Mask R-CNN method is proposed. The proposed method introduces the attention module and multi-scale fusion module to the iris segmentation task. The attention module accelerate the procedure by detecting a smaller iris region for segmentation, while the multi-scale fusion module faithfully preserves the explicit spatial position of iris region. Experimental results on UBIRIS.v2 and CASIA.v4-Distance demonstrate the superior performance of the proposed method.