We propose a method to detect story related subject captions in news video. This paper addresses two issues of caption detection problem. One is the time consuming in the feature computation, the other is the clutter of caption detection results. We first identify the subject caption region based on the frequency of text occurrence. After that, we detect the subject caption beginning frames. In this way, the computation time can be reduced significantly. Meanwhile, the uncorrected types of text are also filtered out, and only the subject caption is detected. Experimental results show that the proposed method can fast and robustly detect subject captions from news video.