Iris center localization is one of the key technologies in attention detection, eye tracking, etc. In order to improve the accuracy of iris center localization for low-resolution images captured under natural light, an iris center localization algorithm based on improved snakuscule is proposed. The algorithm combines the facial landmark detector, snakuscule energy model and iris center quality inspection module. On the basis of the snakuscule, an active radius and an adaptive iteration strategy are adopted, which can ensure the real-time performance and improve the detection accuracy at the same time. Firstly, facial landmarks are used to segment eye ROI. Then, a snakuscule model is initialized based on eye landmarks. During its iteration, either single-pixel iteration or skip iteration is adaptively selected to improve the detection speed of the iris center. Finally, the iris center is inspected to detect whether the forecast center is effective or not. The cascade classifier based on V-J detector is used to re-detect the eyes of the images whose forecast center are ineffective, and then the model will iterate again. The proposed algorithm is tested on BioID dataset. The nomalized error of the proposed algorithm is 85.6%, 96.7% and 99.2% in e d 0.05, e d 0.1 and e d 0.25, respectively. The detection accuracy reaches the mainstream level, which verifies the effectiveness and superiority of the proposed algorithm.