A feasible method combining the characteristics of marine visible image is proposed to detect sea-line in the sequential images from surface vehicle. It is not only appropriate for sea-sky background but also for offshore background. The complexity of sub-images and the average gray difference of their up and down neighborhoods are measured to predict the sea-line region and the consequent processing of images without the existent of sea-line region is given up. Since the sea-line is the longest line with best continuity in the whole nature vision, improved Canny edge detection with surround texture suppression is applied to extract the contour of the object ready for line detection. Weighted vote in Hough transforming is introduced to pick the right line which is horizontal or tilted. The experimental results prove that this method can locate the sea-line region fast and obtain the binary image including the necessary information and attenuating meaningless information. Sea line can be found precisely in the contour edge. It is robust and real-time and is competent for real task where the correct sea-line location is needed.