Image quality assessment and performance evaluation for multimodal biometric recognition using face and iris

Multimodal biometrics offer superior matching performance over a single biometric. However real-life applications increasingly deploy biometric matching systems in non-ideal conditions where good image quality cannot be assured. We assess the impact of image quality on multimodal biometric recognition performance. Biometric recognition of face and iris is carried out on images while the image quality is varied. Image quality is controlled synthetically using defocus, motion blur, light intensity, contrast, resolution and color reduction. Results show that the chosen algorithm for iris recognition is highly sensitive to contrast while brightness has less influence. Effect of defocusing and motion blur on the performance is linear. The effect of pixel resolution is log-linear whereas color reduction has more pronounced effect at extreme values only. The face recognition algorithm is robust to the same variations even though the corresponding image quality followed the trends similar to that for the iris images. For face matching, quality was not a predictor of performance.