Reading text from photographs and videos is a challenging problem that has received a significant amount of attention in recent years. In this work, we propose an approach for single image text detection and we extend this approach to videos. For text detection task, we transform our image into its stroke width representation to extract the basic letter candidates. Then we filter out these candidates using some geometric reasoning methods. For video text detection task, we extract the text locations via same method in each frame. Then, we analyze the motion patterns of the text components in frames to localize video texts. We evaluate our system on ICDAR challenge datasets. Our proposed single frame image text detector as the base of our video text detection approach shows promising results.