In Video based Iris Recognition system, Iris images are obtained with non-Cooperative situation. In video based system, it is time consuming to process each video frame for iris segmentation because some video frames may not contain an eye at all. The raw video frames were not processed like the still images. To improve the efficiency of iris recognition system it is important to process the video frames. Processing of video sequence is more challenging. It requires a robust and accurate method to extract frames that have only iris images. For non-cooperative users there could be many noisy video frames with different quality of iris images. Noisy video frames include blinking eye and blurred frames. In our paper we mainly concentrate on detecting these blink and blurred frames and the results are shown. The results are more accurate than the existing ones.