研究目的
To investigate whether existing no-reference image quality metrics (IQMs) designed for natural images can assess the quality of visible wavelength (VW) iris images based on the biometric system performance.
研究成果
The study concludes that it is possible to use existing no-reference IQMs to assess the quality of VW iris images, with some IQMs showing better performance than others. Re-training IQMs on iris databases can further improve their performance. These findings can be used for the development of robust quality metrics for VW iris image quality and multiple biometric modalities image quality assessment.
研究不足
The study is limited to evaluating no-reference IQMs on VW iris images and does not consider modality-based attributes. The performance of IQMs may be affected by the database used for training.
1：Experimental Design and Method Selection:
The study evaluates the performance of 15 selected no-reference IQMs on VW iris biometrics using a near infrared iris recognition algorithm adapted to VW iris samples.
2：Sample Selection and Data Sources:
Iris images from the GC2 multimodality biometric database are used, with artificially degraded iris images to simulate various distortions.
3：List of Experimental Equipment and Materials:
The database includes images acquired with a Lytro first generation Light Field Camera, a Google Nexus 5 embedded camera, and a Canon D700 with Canon EF 100mm f/
4：8L Macro Lens. Experimental Procedures and Operational Workflow:
The study introduces four types of distortions related to image-based quality attributes and evaluates the performance of IQMs using DET curves and EER values.
5：Data Analysis Methods:
The performance of IQMs is assessed based on their ability to predict iris recognition algorithm performance, with some IQMs re-trained on VW iris images to improve their performance.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
